Meeting Title: Brainforge x CTA: Weekly! Date: 2026-03-13 Meeting participants: Awaish Kumar, Katherine Bayless, Chi Quinn, Kyle Wandel, Uttam Kumaran, Ashwini Sharma
WEBVTT
1 00:00:30.020 ⇒ 00:00:31.890 Katherine Bayless: Ao, good morning.
2 00:00:32.850 ⇒ 00:00:34.460 Awaish Kumar: Hi, good morning.
3 00:00:37.280 ⇒ 00:00:38.290 Katherine Bayless: Alright.
4 00:00:42.890 ⇒ 00:00:43.990 Awaish Kumar: Hello, you’re doing?
5 00:00:44.890 ⇒ 00:00:46.010 Katherine Bayless: Good, yeah.
6 00:00:46.230 ⇒ 00:00:47.940 Katherine Bayless: Happy it’s Friday.
7 00:00:49.690 ⇒ 00:00:51.320 Awaish Kumar: Do you have a question?
8 00:00:51.970 ⇒ 00:00:52.690 Katherine Bayless: Sorry?
9 00:00:52.900 ⇒ 00:00:54.779 Awaish Kumar: Do you have any plans for the weekend?
10 00:00:55.220 ⇒ 00:00:59.919 Katherine Bayless: Hmm, not really. We’ll see what comes up. What about you?
11 00:01:01.340 ⇒ 00:01:05.140 Awaish Kumar: Yeah, I’m… I’ll be watching maybe just Netflix or something.
12 00:01:05.580 ⇒ 00:01:07.610 Katherine Bayless: Sounds good.
13 00:01:07.610 ⇒ 00:01:13.860 Katherine Bayless: Actually, I watched the new, documentary on HBO about Fukushima, the other day, and it was…
14 00:01:13.860 ⇒ 00:01:30.829 Katherine Bayless: It’s very interesting. Like, I definitely did not know as much about the incident, as the documentary kind of got into, so it was… I mean, that’s my happy place, is, like, learning new nerdy things. And then the, like, follow-up one, Kai, I think you probably saw in the Slack channel that Courtney mentioned, that one was really good, too.
15 00:01:30.830 ⇒ 00:01:40.430 Chi Quinn: Yeah, that’s my… that’s my plan for this weekend, to watch that, because I’m so curious to hear about it, so… Yeah. No spoilers!
16 00:01:40.790 ⇒ 00:01:42.950 Chi Quinn: No spoilers.
17 00:01:43.230 ⇒ 00:01:46.510 Katherine Bayless: Rosebud is the sled. No.
18 00:01:47.380 ⇒ 00:01:49.620 Katherine Bayless: What is it called? What is it called?
19 00:01:50.000 ⇒ 00:01:53.320 Katherine Bayless: I think it’s just called Fukushima?
20 00:01:53.320 ⇒ 00:01:54.010 Chi Quinn: Yeah.
21 00:01:54.350 ⇒ 00:01:58.219 Katherine Bayless: I think so. Yeah, I don’t think it had, like, any sort of further subtitle.
22 00:01:59.970 ⇒ 00:02:01.250 Uttam Kumaran: Is that a new show?
23 00:02:01.720 ⇒ 00:02:04.379 Katherine Bayless: Oh, there’s a new documentary on HBO.
24 00:02:04.380 ⇒ 00:02:07.740 Uttam Kumaran: Oh, okay. Is it good? I saw the, like, tile for it.
25 00:02:08.039 ⇒ 00:02:16.099 Katherine Bayless: It was. I mean, I like that kind of stuff, right? Like, I’m nerdy, like that, but, I thought it was really interesting. It definitely gets into the, like…
26 00:02:16.189 ⇒ 00:02:28.139 Katherine Bayless: the, like, on-the-ground reality of the TEPCO employees in a way that I don’t think the media coverage at the time really did. And then there was, like, a lot of science-y stuff that I definitely learned.
27 00:02:29.710 ⇒ 00:02:34.449 Uttam Kumaran: There’s… what was the, oh, the… what was the other nuclear fallout show? It was a Chernobyl?
28 00:02:35.790 ⇒ 00:02:37.020 Uttam Kumaran: That was really, really good.
29 00:02:37.250 ⇒ 00:02:40.880 Katherine Bayless: Yeah, yeah, I still haven’t actually watched that show.
30 00:02:40.880 ⇒ 00:02:47.600 Uttam Kumaran: Really? Oh, you’d love it. It’s like a… it’s drama… But it’s So stressful.
31 00:02:47.600 ⇒ 00:02:48.780 Katherine Bayless: Yeah, yeah.
32 00:02:48.780 ⇒ 00:02:49.630 Uttam Kumaran: Yeah.
33 00:02:50.070 ⇒ 00:03:04.619 Katherine Bayless: But the, the person, or one of the folks that we work with pretty closely on the marketing team is, like, a fanatic for, like, nuclear fallout, apocalypse-type things, and so I posted in our Slack channel, I said, please come prepared to discuss on Monday.
34 00:03:04.620 ⇒ 00:03:06.220 Uttam Kumaran: Nice.
35 00:03:06.220 ⇒ 00:03:07.429 Kyle Wandel: I mean, who is that?
36 00:03:07.880 ⇒ 00:03:08.640 Katherine Bayless: Courtney?
37 00:03:08.850 ⇒ 00:03:10.430 Kyle Wandel: Oh, okay. Interesting.
38 00:03:11.420 ⇒ 00:03:11.750 Katherine Bayless: Yeah.
39 00:03:11.750 ⇒ 00:03:13.240 Kyle Wandel: Yeah, she seems that way.
40 00:03:13.480 ⇒ 00:03:16.640 Katherine Bayless: It’s a delightful surprise, Quirk.
41 00:03:18.530 ⇒ 00:03:19.470 Kyle Wandel: Nice.
42 00:03:19.760 ⇒ 00:03:20.790 Uttam Kumaran: How’s the rest of the week?
43 00:03:22.850 ⇒ 00:03:29.040 Katherine Bayless: Good, eventful, a little chaotic. I think, finally, today I might get a chance to, like.
44 00:03:29.350 ⇒ 00:03:33.659 Katherine Bayless: dig out a little bit, but but not bad. Just busy, hectic.
45 00:03:33.660 ⇒ 00:03:34.290 Uttam Kumaran: Okay.
46 00:03:35.100 ⇒ 00:03:36.140 Katherine Bayless: Was that for you guys.
47 00:03:37.640 ⇒ 00:03:44.240 Uttam Kumaran: Good, I think always busy. I feel like I’ve made some good progress, and, like, I kind of now understand,
48 00:03:44.520 ⇒ 00:03:47.379 Uttam Kumaran: sort of semantic views, end-to-end, I have some…
49 00:03:47.510 ⇒ 00:03:52.140 Uttam Kumaran: Couple things to share, but I think we also made a lot of progress on the modeling side.
50 00:03:52.340 ⇒ 00:03:53.470 Uttam Kumaran: Yep.
51 00:03:54.030 ⇒ 00:03:55.880 Uttam Kumaran: So, maybe we can…
52 00:03:56.260 ⇒ 00:04:03.120 Uttam Kumaran: Probably start there, and then we’re working today… working yesterday and today on, like, some of the identity stitching models as well.
53 00:04:03.490 ⇒ 00:04:09.220 Katherine Bayless: Yeah, actually, before we dive into the QA, just a couple sort of informational pieces,
54 00:04:09.380 ⇒ 00:04:22.140 Katherine Bayless: So one piece of good news we got this week, was that the board did approve the additional funding request to replace the AWS ProServe contract, so we will have that money back in our budget, which is awesome.
55 00:04:22.140 ⇒ 00:04:46.899 Katherine Bayless: That work will start on Monday with the kickoff call. It’ll take about probably 8 weeks or so, probably 10, if we’re being honest. But Kyle and I, along with Jay and probably Ian from Jay’s team, will be taking lead on getting that work, sort of shepherded across the finish line. I think we’re probably also gonna loop in the market research folks, if they have the bandwidth to kind of
56 00:04:46.900 ⇒ 00:05:01.200 Katherine Bayless: come along for the ride, because they’re still using a bunch of old SQL servers that are hosted in Azure, and I think this is a logical time for us to just at least lift and shift those into AWS, and that way, you know, our cloud footprint is a little bit more consolidated.
57 00:05:01.220 ⇒ 00:05:15.670 Katherine Bayless: So yeah, so that work will kick off on Monday. Keep you guys posted in terms of, like, when we have the new accounts, and if we’re gonna sort of detach the ones we have and attach them to the new structure, or just build them out from scratch, which I’m hoping we will go with the latter.
58 00:05:18.210 ⇒ 00:05:28.470 Katherine Bayless: Then the other informational thing I had walked out of my brain. That one will come back. Anyway, okay, so the identity stitching was what kind of jogged my memory here, so…
59 00:05:28.470 ⇒ 00:05:42.730 Katherine Bayless: I did some work yesterday to kind of figure out, like, a little bit of a release, so to speak, timing for us for the next two quarters, so at the end of March, we’ll get the audited data back for CES, and so that
60 00:05:42.730 ⇒ 00:05:59.020 Katherine Bayless: is kind of a natural, good time for us to say, like, the audited data is ready, as is Snowflake, for those who are interested, ready and willing, like, you can come into Snowflake this way, we’ll set up some office hours, do a little bit of training, but it just gives us a kind of a really, like.
61 00:05:59.020 ⇒ 00:06:04.609 Katherine Bayless: easy, logical way to do a, like, initial rollout of Snowflake to the broader organization, and kind of…
62 00:06:04.650 ⇒ 00:06:25.480 Katherine Bayless: announce more officially that Power BI is going away, and that kind of stuff. So, I think we’re probably on perfect track to get there, but yeah, delivering all of this stuff by the end of March, I think is a good end of Q1 release, and also getting the new audited data file will be a good test of flowing it all through to the final, sort of, dbt, structures.
63 00:06:25.500 ⇒ 00:06:33.389 Katherine Bayless: Because we’ll be doing that on a daily basis once we get to September, so this is at least one early way to test.
64 00:06:33.510 ⇒ 00:06:35.540 Katherine Bayless: And then sort of the… yeah.
65 00:06:35.540 ⇒ 00:06:46.089 Awaish Kumar: I have a question here, like, when we get the new files, what, like, what do you want, like, should we just replace our sources for, or do we want to, like.
66 00:06:46.220 ⇒ 00:06:48.210 Awaish Kumar: To keep the version of the data?
67 00:06:48.870 ⇒ 00:07:11.669 Katherine Bayless: I think in terms of what we keep in Snowflake, we want to just keep the most recent sort of file, so it would be the audited one. In the data lake, I do think we will keep a backup of the pre-audit and pre-pre-audit? Post-show? Post-show, pre-audit, and audit are kind of the three flavors. So we’ll probably keep the post-show and the pre-audit reports in the data lake, just in a, you know, archive bucket.
68 00:07:11.670 ⇒ 00:07:15.420 Katherine Bayless: But the latest version of the data should always be what’s in Snowflake.
69 00:07:16.630 ⇒ 00:07:17.220 Awaish Kumar: Okay.
70 00:07:18.790 ⇒ 00:07:28.040 Katherine Bayless: And then, so, once we’re kind of done with that, the next big, I think, logical sort of release time for us will be…
71 00:07:28.210 ⇒ 00:07:44.410 Katherine Bayless: End of June, conveniently also end of Q2, because July 1st is when our Innovation Awards launch, and that’s gonna be our first chance to really test, you know, for reals, our ability to identity stitch exhibitors and member companies from those two different systems.
72 00:07:44.410 ⇒ 00:08:00.059 Katherine Bayless: Because the Innovation Awards, there’s, like, a pricing incentive for if you are an exhibitor and a member, and in order to deliver that, pricing, we need to know. And so last year, the platform, the oper… The, the, what’s the word I’m looking for?
73 00:08:00.630 ⇒ 00:08:25.580 Katherine Bayless: the means of uploading the data was, like, two CSVs on the back end. Kyle’s doing some legwork to figure out if they have APIs. It might still wind up being two CSVs, to be honest. I’m not sure if Open Water has APIs, but hopefully they do. And even if it still has to be two CSVs, at least we’ll be able to have the data in parity across them, unlike last year, where it was just kind of chaos. So yeah, so the identity stitching sort of V0, being ready by the end of June.
74 00:08:25.580 ⇒ 00:08:30.550 Katherine Bayless: So that we can support that process more robustly than in the past is perfect.
75 00:08:30.550 ⇒ 00:08:55.130 Katherine Bayless: And it nicely kind of gives us a little bit of a lower stakes way, because the volume in that program isn’t massive, to start testing out that identity stitching before we get to the fall, and it becomes a little bit more hectic with reporting requests, looking across exhibitors and registrants and that kind of thing, so… So yeah, so end of March and end of June, I think delivering the CES data and then delivering the V0 identity stitching is a good, sort of.
76 00:08:55.380 ⇒ 00:09:05.290 Katherine Bayless: Those are, like, the org-wide relevant releases that we’ll have, and then we’ll obviously have smaller things for the membership and marketing teams along the way, but those are the kind of two big dates to target.
77 00:09:05.990 ⇒ 00:09:06.750 Awaish Kumar: Okay.
78 00:09:07.190 ⇒ 00:09:10.659 Awaish Kumar: I have a few questions on identity strategy, like, we are…
79 00:09:10.960 ⇒ 00:09:19.819 Awaish Kumar: like, like, last time I worked with an exhibitor space, data, which is, and then there we have a CES data.
80 00:09:20.170 ⇒ 00:09:25.350 Awaish Kumar: And then we have some data from remembers, where we have this team organization.
81 00:09:25.520 ⇒ 00:09:28.460 Awaish Kumar: So, basically, like.
82 00:09:28.750 ⇒ 00:09:36.289 Awaish Kumar: what is it? Like, I want to clarify the scope, like, we have… right now, most of our work is being done on, for example, registration data.
83 00:09:37.460 ⇒ 00:09:54.859 Awaish Kumar: For the registration data, for each individual registrant, we have company names. So we are… we have created DIMCES Company, as Kyle, I’ve already seen that. So one way is that we have this DIMS company table from CES data, and we try to match this data with the
84 00:09:55.160 ⇒ 00:10:03.199 Awaish Kumar: With the data in the… In the, like, the… Like, remembers, and
85 00:10:03.380 ⇒ 00:10:08.599 Awaish Kumar: and get the… try to get the Impexium ID, or organization ID.
86 00:10:09.230 ⇒ 00:10:23.759 Katherine Bayless: - actually, Kyle, sorry, no. In this case, this is the data that Kyle just delivered this morning. So, the exhibitor data from ExpoCAD is the data that we’re going to need for the identity stitching to support the Innovation Awards.
87 00:10:23.880 ⇒ 00:10:39.290 Katherine Bayless: So we’ll be getting that raw export from ExpoCAD. Kyle and I will start rebuilding that pipeline, because they are selling exhibit space now, so we can start consuming that more regularly. But yeah, the, companies that are in that, booths
88 00:10:39.290 ⇒ 00:10:46.899 Katherine Bayless: file in the ExpoCAD data, those are the ones that we’re going to need to be able to tie to the member companies from remembers.
89 00:10:47.060 ⇒ 00:10:58.289 Katherine Bayless: CES registration also needs to be joinable into the identity stitching, but it’s the exhibitor data from ExpoCAD and the company, or the membership data from remembers that are the two pieces for the Innovation Awards.
90 00:11:00.540 ⇒ 00:11:03.610 Awaish Kumar: Okay, yeah.
91 00:11:03.860 ⇒ 00:11:06.099 Awaish Kumar: And basically… This is exciting.
92 00:11:06.370 ⇒ 00:11:19.860 Kyle Wandel: Sorry, the matching table is correct, everything is white, but I think you are also right, Ush, in terms of, like, we still want to add all of the companies that are listed in the registration table as, like, a…
93 00:11:20.300 ⇒ 00:11:36.519 Kyle Wandel: whatever, user selected or user input table, and then that needs to match to something, whether it be remembered… well, it needs to match to members and ExpoCAD and Salesforce. I think I’m… I don’t know if I’m missing anything but else, but I think those are the three main databases that we have.
94 00:11:37.240 ⇒ 00:11:37.950 Awaish Kumar: Okay.
95 00:11:38.650 ⇒ 00:11:57.220 Kyle Wandel: So it’s like taking… well, and then maybe this is oversimplifying it, Catherine, I don’t know if you want to break it out even more, but it’s, like, almost taking, like, what is every single company name in DIM organizations, DIM CES company, and then DIM Exhibitor booth, and then mashing that to the three…
96 00:11:58.590 ⇒ 00:12:01.169 Kyle Wandel: Foreign keys, or primary keys from the main systems.
97 00:12:02.770 ⇒ 00:12:15.319 Katherine Bayless: Yeah, I mean, I think that’s where we ultimately do want to get to. I think… so if we kind of think about the, like, the waterfall of matching ability, so if we’re starting with the ExpoCAD data and the Remembers data.
98 00:12:15.320 ⇒ 00:12:26.090 Katherine Bayless: were coming from, sort of, more constrained record sets, where it’s, you know, this is a record for that company in that system with an identifier, right? The granularity is the company itself.
99 00:12:26.290 ⇒ 00:12:42.830 Katherine Bayless: Then in, like, the registration data and all of the other places where we let a human tell us what company they work for, that’s all of that, like, really messy, chaotic data. And so if we can get parity, which I think this part will be fairly easy, because the two teams have tried to keep track of each other.
100 00:12:42.830 ⇒ 00:12:47.919 Katherine Bayless: If we can get the parity between the exhibitor data and the remembers data.
101 00:12:47.920 ⇒ 00:13:11.649 Katherine Bayless: then the next phase of matching is like, okay, all of the chaos that Kyle talked about that is self-entered, we need to match that up also with those companies, and that’s where we’ll use that, what is it? Alias and domain table in remembers, so as we find, you know, flavors, permutations of company names that we do want to associate with a canonical company name.
102 00:13:11.650 ⇒ 00:13:17.060 Katherine Bayless: We can add that alias information into remembers so that it’ll flow through as a matchable row.
103 00:13:18.510 ⇒ 00:13:34.049 Kyle Wandel: So then for, like, a priority list, Catherine, correct me if I’m wrong, we should… they should focus on their exhibitor… tie in exhibitor booth, the member first, and then registration table at the, whatever, the identities… start trying to match the identities listed in CES registration. Okay.
104 00:13:34.200 ⇒ 00:13:35.890 Katherine Bayless: Yeah, exactly.
105 00:13:35.890 ⇒ 00:13:42.230 Awaish Kumar: And… Yeah, for the exhibitor data, I see there’s only one, like, exhibit space.
106 00:13:42.720 ⇒ 00:13:54.850 Kyle Wandel: Yeah, so I just added all of that data lake that Aswini’s gonna add and take a look at, and that is all the exhibitor data that we have as of its rawest form, so it goes back to 21. 21 is
107 00:13:54.850 ⇒ 00:14:03.709 Kyle Wandel: weird because it’s digital, so some of the files aren’t even necessary, so, like, hotel rooms or badge allocation. But everything else is there from 2021.
108 00:14:03.820 ⇒ 00:14:09.360 Kyle Wandel: We can probably try and dig around for more, but that’s all that we have, or that we know of right now.
109 00:14:10.090 ⇒ 00:14:12.230 Katherine Bayless: Yeah. It’s plenty to start with.
110 00:14:12.230 ⇒ 00:14:13.469 Kyle Wandel: Yeah, I figured so.
111 00:14:16.540 ⇒ 00:14:20.339 Awaish Kumar: Yeah, okay, yeah, we can, then take a…
112 00:14:20.460 ⇒ 00:14:23.369 Awaish Kumar: stab on that first. We are already…
113 00:14:24.090 ⇒ 00:14:30.759 Awaish Kumar: Started with the exhibited space table we had, but now, since there’s more data, we can look into that.
114 00:14:31.130 ⇒ 00:14:49.260 Kyle Wandel: Yeah, I would imagine that any exhibitor data, I mean, I mean, maybe correct me if I’m wrong, I would imagine that just ignore any exhibitor data before this, and then just focus on the ones that are those JSON files that are in the S3 now, because those are the rawest form of it, and, like, I guess the best form, so…
115 00:14:50.080 ⇒ 00:14:52.489 Katherine Bayless: And that’s what we’ll have on a going-forward basis, yeah.
116 00:14:52.750 ⇒ 00:14:53.230 Katherine Bayless: Yeah.
117 00:14:53.230 ⇒ 00:14:53.820 Kyle Wandel: Yeah.
118 00:14:54.760 ⇒ 00:15:06.480 Kyle Wandel: So whatever’s in remembers is not correct. I know that they have created a couple things in Remembers, like Exhibit Booth, but that’s not… that’s not still… that doesn’t store our exhibitor data. Maybe one day it could, but as of right now, it doesn’t.
119 00:15:06.950 ⇒ 00:15:20.529 Katherine Bayless: Yeah. Yeah, I mean, I think once we have the identity stitching and all of these integrations in place, then we can absolutely start putting correct data into remembers around that. Like, I do look forward to being able to help them actually store some of this information in that system.
120 00:15:22.190 ⇒ 00:15:22.910 Awaish Kumar: Okay.
121 00:15:23.210 ⇒ 00:15:27.790 Awaish Kumar: And on the other side, I worked on the CES modeling.
122 00:15:28.110 ⇒ 00:15:32.239 Awaish Kumar: Kyle, I’ve already seen some of it, I have pushed a few more changes to it.
123 00:15:33.950 ⇒ 00:15:46.060 Awaish Kumar: And, like, the data exactly matches with what we… like, the version we had before, but I… yeah, we have to, like… I see… I saw your comment, Carl, so…
124 00:15:46.390 ⇒ 00:15:50.989 Awaish Kumar: Maybe do one more pass on… on that, and
125 00:15:51.850 ⇒ 00:15:57.590 Awaish Kumar: Yeah, like, I’m basically trying to… now that we are dividing each…
126 00:15:57.820 ⇒ 00:16:00.860 Awaish Kumar: The columns as a subset, like, for example.
127 00:16:01.280 ⇒ 00:16:05.410 Awaish Kumar: That these columns belong to exhibitor, DM exhibitor.
128 00:16:05.620 ⇒ 00:16:20.420 Awaish Kumar: table, and then what should be the key, right? Exhibitor ID might be the primary key, so I’m just trying to find out those things, and some of the tables, I’m not able to figure that out, what should be the key, and I’m just using a hash for now.
129 00:16:21.840 ⇒ 00:16:25.869 Awaish Kumar: But yeah, we can, like, obviously trade on that and continue to improve.
130 00:16:25.870 ⇒ 00:16:43.070 Kyle Wandel: Yeah, I saw that you were doing some stuff this morning, so I was putting that stuff for Sweeney together this morning, and then I’ll take a look at the QA and then make comments. I know that I did notice that you made a lot of changes already, so thank you. I think there are a couple more things that I noticed that I was gonna comment on, so I’ll get that to you today.
131 00:16:45.090 ⇒ 00:16:53.459 Katherine Bayless: I also was able to start kind of going through this morning and taking a look at some of the validation queries, and I think I found a few question marks,
132 00:16:53.590 ⇒ 00:16:58.979 Katherine Bayless: So, I can put my comments in the PR as well, along with Kyle’s, if that makes the most sense.
133 00:16:59.220 ⇒ 00:17:00.439 Awaish Kumar: Okay, yeah, sure.
134 00:17:01.130 ⇒ 00:17:11.830 Katherine Bayless: I did have one really silly question. In the PR, you mentioned that you had created, docs slash ces2026 star schemaelMap.markdown.
135 00:17:11.960 ⇒ 00:17:13.329 Katherine Bayless: But I can’t find it.
136 00:17:13.510 ⇒ 00:17:16.319 Katherine Bayless: Where… am I looking in the wrong place? I probably am.
137 00:17:16.329 ⇒ 00:17:21.299 Awaish Kumar: Well, actually, it just… The description got pushed by the document didn’t.
138 00:17:21.710 ⇒ 00:17:22.949 Katherine Bayless: Oh, God.
139 00:17:23.959 ⇒ 00:17:29.119 Awaish Kumar: I just have it, like, with me, a shorter version of that.
140 00:17:29.729 ⇒ 00:17:37.209 Awaish Kumar: so it does not exactly have the field-level mapping right now, because I cut it out for this.
141 00:17:37.579 ⇒ 00:17:41.219 Awaish Kumar: meeting. But it’s, it has, like,
142 00:17:41.689 ⇒ 00:17:45.579 Awaish Kumar: What, like, this is the base… Table now that we are…
143 00:17:45.829 ⇒ 00:17:50.029 Awaish Kumar: creating it on top of it. This is the registration table.
144 00:17:50.399 ⇒ 00:17:52.679 Awaish Kumar: That is our…
145 00:17:52.909 ⇒ 00:18:01.169 Awaish Kumar: like, transac- what you say, the individual transactions for, like, the… it has registration key, and the edit gen…
146 00:18:02.059 ⇒ 00:18:19.519 Awaish Kumar: And that… that is a combination of a year and a consistent ID. Then it joins to the person, and it joins back to the company, it joins back to the exhibitor data using foreign keys, and then it has some flags, like, if it… it was somebody attended that, they just…
147 00:18:19.739 ⇒ 00:18:25.509 Awaish Kumar: That registrant attended the… Or not, like, if there’s a flag for that.
148 00:18:25.699 ⇒ 00:18:30.469 Awaish Kumar: And, things like that. So, if we want to add any measures, then we can add here.
149 00:18:30.709 ⇒ 00:18:33.389 Awaish Kumar: And then it joins back,
150 00:18:33.969 ⇒ 00:18:37.149 Awaish Kumar: with all these different DIMM models which are here.
151 00:18:37.319 ⇒ 00:18:41.349 Awaish Kumar: No, I’m like… CS person, name company, registration…
152 00:18:41.950 ⇒ 00:18:53.880 Katherine Bayless: Yeah, okay, so I was… this was actually why I was looking for the document, because I was curious… I couldn’t figure out why we had backed CES registration and DIM CES registration, because
153 00:18:53.930 ⇒ 00:19:06.630 Katherine Bayless: I was… so my initial, you know, going through the validation queries, I was trying to make sure, at least, you know, at the top line, we get the 148.406, or whatever it is, for attendees, and it doesn’t come back that way.
154 00:19:06.830 ⇒ 00:19:26.039 Katherine Bayless: But it… and I think that has a different piece of feedback, but I also noticed it wasn’t… you couldn’t go straight from FACT CES registration to filtering on the attendee type, or the registration type, without going from FACT to DIMM to the other DIM registration type. And so I was like, if we’re going to have to hit both almost every time, should we do this differently?
155 00:19:26.910 ⇒ 00:19:32.069 Awaish Kumar: No, I have made few changes after that. You can directly, use,
156 00:19:32.450 ⇒ 00:19:36.910 Awaish Kumar: Like, there’s no registration type, dim now, I think, yeah.
157 00:19:37.430 ⇒ 00:19:41.100 Awaish Kumar: I’ve removed that, and it is in the… as part of the…
158 00:19:41.270 ⇒ 00:19:46.319 Awaish Kumar: the fact table itself, but then why I have these two is because,
159 00:19:46.750 ⇒ 00:19:55.610 Awaish Kumar: like, there’s a lot of fields that are basically related… that does not belong to… Any of the…
160 00:19:57.650 ⇒ 00:20:04.609 Awaish Kumar: like, the dim tables we have, like, they are for the registration. For example, if registration was canceled, or something like that.
161 00:20:04.730 ⇒ 00:20:15.979 Awaish Kumar: So, I just wanted to shorten this table, so it has, for example, the exact fields that are meaningful to join with other team tables and get the answers we need.
162 00:20:16.180 ⇒ 00:20:21.299 Awaish Kumar: But there’s a lot of information regarding registration itself, right?
163 00:20:21.370 ⇒ 00:20:37.510 Awaish Kumar: that lives there. So, there is no difference in those tables, because they are joining on a registration key. So, basically, this DIM is basically an extension of that fact. We can include all those fields in the fact as well, and remove this.
164 00:20:38.970 ⇒ 00:20:59.289 Katherine Bayless: Okay, okay, I’ll take a look at it again, then, with that understanding of it, because it might make sense to combine them, but it also might make sense to just add the registration type to the fact table, but I can… I can take a look at them again, now that I understand the intent a little better. I think the other pieces that were affecting the count were just smaller things. It looked like the…
165 00:20:59.290 ⇒ 00:21:05.690 Katherine Bayless: is student field was null for everybody, but there should be about 400 or so records, I think.
166 00:21:05.690 ⇒ 00:21:10.649 Katherine Bayless: That have that flag, so it might have just gotten lost in translation somewhere along the way.
167 00:21:10.690 ⇒ 00:21:13.000 Katherine Bayless: And then the other one I had was the.
168 00:21:13.000 ⇒ 00:21:14.090 Awaish Kumar: I think it has…
169 00:21:14.430 ⇒ 00:21:17.660 Kyle Wandel: We have it filtered out in the base registration table, I believe.
170 00:21:18.390 ⇒ 00:21:21.569 Awaish Kumar: I think it… I have fixed that, like, previously.
171 00:21:21.810 ⇒ 00:21:33.529 Awaish Kumar: like, it was all coming in, and I was not filtering on a stone in the… in the final aggregations. So I have added those, in the… in the…
172 00:21:33.870 ⇒ 00:21:38.430 Awaish Kumar: in the table called Report. So basically, then, I’ve already created a…
173 00:21:39.790 ⇒ 00:21:42.859 Awaish Kumar: A table like this, which is basically…
174 00:21:43.010 ⇒ 00:21:51.519 Awaish Kumar: Normally we call these, like, summary tables, which basically joins… so we affect on them, basically, to see our data.
175 00:21:52.170 ⇒ 00:21:53.319 Awaish Kumar: Like that.
176 00:21:53.800 ⇒ 00:21:57.049 Awaish Kumar: For the data integrity, and there’s low redundancy in our data.
177 00:21:57.150 ⇒ 00:21:58.280 Awaish Kumar: But then…
178 00:21:58.620 ⇒ 00:22:10.349 Awaish Kumar: We need to join all of that, and we don’t want to, like, join this data, like, for individual query, maybe? Like, because every time we need to… we know that these are 3-4 tables that needs to be joined.
179 00:22:10.560 ⇒ 00:22:14.349 Awaish Kumar: Using those, we can just create this summary, table.
180 00:22:14.460 ⇒ 00:22:25.940 Awaish Kumar: Which basically already joins, like, beforehand. So when you want to curate, for somebody else, they can just use this table instead of managing individual joins.
181 00:22:26.160 ⇒ 00:22:35.509 Uttam Kumaran: I think one of the ways to also, like, come at it from, like, the adoption side is we try to just, like, say, okay, these tables are always gonna get joined together.
182 00:22:35.640 ⇒ 00:22:38.190 Katherine Bayless: So we might as well, like, handle some of that.
183 00:22:38.240 ⇒ 00:22:42.040 Uttam Kumaran: But, yes, it will… you can facilitate it in both.
184 00:22:42.180 ⇒ 00:22:49.859 Uttam Kumaran: Ultimately, once we start to see how people are joining these things together, it allows us to create those helpful, you know, like, report tables.
185 00:22:50.000 ⇒ 00:22:53.540 Uttam Kumaran: But, like, it’s still actually really helpful to maintain
186 00:22:53.750 ⇒ 00:22:59.409 Uttam Kumaran: the modular pieces. The alternative world is report-specific
187 00:22:59.950 ⇒ 00:23:16.539 Uttam Kumaran: fat model tables, right? So, it’s sort of like thinking through the best of both worlds. I know it’s a little bit confusing, so one… on our side, like, we could… if it’s just us, we could pause on creating any of the report tables until we get a sense of, like, what they could be, but we sort of just…
188 00:23:16.730 ⇒ 00:23:18.920 Uttam Kumaran: Took some liberty to create a few of those.
189 00:23:19.760 ⇒ 00:23:24.090 Katherine Bayless: Yeah, no, that makes a lot of sense, to be honest. Like I said, I think I was…
190 00:23:24.330 ⇒ 00:23:36.099 Katherine Bayless: without that… this document, I’m struggling to kind of put the pieces together, but that makes sense to have some of the reports that are most common, kind of, like, preset, because there are definitely a handful we know that
191 00:23:36.100 ⇒ 00:23:45.560 Katherine Bayless: it makes sense to do that with. I think the question, more of a curiosity then, kind of becomes, maybe for you, Utam, and the work you’ve been doing around the semantic view stuff, is like.
192 00:23:45.560 ⇒ 00:24:00.780 Katherine Bayless: I wonder, like, which one Cortex code will gravitate towards, because probably, with the exception of the things that we will park in, like, dashboards, a lot of the querying will be done by the robot, because a person will ask a question, and so would we, I guess, maybe want to guide… actually.
193 00:24:00.840 ⇒ 00:24:06.830 Katherine Bayless: Small, important side note, I’ve learned internally at Snowflake, they call Cortex code COCO.
194 00:24:06.830 ⇒ 00:24:08.109 Uttam Kumaran: Okay. Would…
195 00:24:08.110 ⇒ 00:24:15.380 Katherine Bayless: know to, like, use the report table versus the fact in DIM tables, would we want it to have a preference,
196 00:24:15.380 ⇒ 00:24:18.110 Uttam Kumaran: Yeah, so I’m gonna kind of present today a little bit on, like.
197 00:24:18.350 ⇒ 00:24:22.440 Uttam Kumaran: Where to add context, but yes, like.
198 00:24:22.570 ⇒ 00:24:31.780 Uttam Kumaran: there’s gonna be some context that’s like, oh, this is this metric, but then there’s gonna be some that’s gonna be… I don’t know, it’s more about, like, us steering.
199 00:24:31.890 ⇒ 00:24:39.789 Uttam Kumaran: And so, we can put that into, like, into custom instructions. There’s a concept of, like, verified queries.
200 00:24:39.890 ⇒ 00:24:46.040 Uttam Kumaran: Like, queries that we know are, like, the ways to answer certain questions,
201 00:24:46.390 ⇒ 00:24:49.590 Uttam Kumaran: So, there’s gonna be a few ways to steer. I mean, ultimately, what I’m…
202 00:24:49.950 ⇒ 00:25:07.190 Uttam Kumaran: trying to do more of is, like, we predict the questions, and then we constantly test whether it’s answering in the way that we want it to. Then we… then we can tweak, but yes, like, ideally, you’re right, like, nobody in the team… nobody in other users necessarily need to know the components.
203 00:25:07.610 ⇒ 00:25:18.440 Uttam Kumaran: it would be great. Either way, the goal is ultimately answering their question, but also, I was talking to, you know, someone yesterday about, oftentimes, someone comes in with a question, but
204 00:25:18.480 ⇒ 00:25:28.669 Uttam Kumaran: it’s actually, like, they may have a specific question, but frankly, they actually more have, like, a topic that they want to discuss. For example, like, I’m interested in, like.
205 00:25:28.730 ⇒ 00:25:31.139 Uttam Kumaran: the CES batch scans, like…
206 00:25:31.390 ⇒ 00:25:48.889 Uttam Kumaran: growth over time, but… or they… but they may come in and be like, I want to see badge scans, this, this, this. So I’m trying to… I think one thing that’ll be interesting is, like, can we train people to come more with, like, a topic they want to, like, discuss? Because the AI can then go to multiple places and piece it together, versus, like.
207 00:25:49.160 ⇒ 00:25:51.930 Uttam Kumaran: It’ll just give you the badge scan data, but…
208 00:25:52.370 ⇒ 00:25:55.670 Uttam Kumaran: Then have to go do the work of piecing it together, so…
209 00:25:56.070 ⇒ 00:25:56.580 Katherine Bayless: I like that.
210 00:25:56.580 ⇒ 00:26:01.189 Uttam Kumaran: Long story short, like, yes, and I think we’ll just have to kind of watch how people use it.
211 00:26:03.290 ⇒ 00:26:04.050 Awaish Kumar: Very good.
212 00:26:04.740 ⇒ 00:26:16.349 Kyle Wandel: from my side, two… just very, two things to note. Do we want to do, like, a DIM CES attendee, or do we want to do it, like, in the semantic model when using Cortex that just says, when…
213 00:26:16.370 ⇒ 00:26:38.630 Kyle Wandel: this is what an attendee is. That’s one big question. And then the other big kind of thing is that instead of using DIMCES year event from the registration table, I created that new S3, like, kind of like, basically lookup table. So I used that as, like, the events table, and that has CES unveiled and CES from, like, the 60s all the way to now. So…
214 00:26:38.630 ⇒ 00:26:44.769 Kyle Wandel: That’ll be a more comprehensive report, rather than just grabbing it from, CS registration.
215 00:26:45.420 ⇒ 00:26:50.949 Awaish Kumar: So, like, that dates… like, the Unveiled, or CS, are these different events, or…
216 00:26:51.360 ⇒ 00:27:00.350 Kyle Wandel: Yeah, they’re slightly different events, so we have… basically, Unveiled is, like, a marketing media event, and then the CES is the actual show.
217 00:27:00.350 ⇒ 00:27:13.559 Kyle Wandel: Usually Unveiled happens, like, 2 days before, and we also do Unveiled in different locations, so, like, around the world. I think this year we’re doing one in Boston, I believe. So there’s a lot of different Unveiled
218 00:27:13.590 ⇒ 00:27:24.419 Kyle Wandel: kind of events that happen. We don’t necessarily have registration for them other than the badge scan codes at CES. I don’t think we have anything else. Do you know, Catherine?
219 00:27:24.980 ⇒ 00:27:35.279 Katherine Bayless: So yeah, I’m realizing as you’re talking. Unveiled is gonna be confusing. So yeah, so there’s the Unveiled event, like Kyle said, that’s part of the CES in Vegas.
220 00:27:35.320 ⇒ 00:27:47.799 Katherine Bayless: all of that registration data is the stuff that’s in all of this star schema. The only way to know whether or not somebody attended CES and Unveiled, both in Las Vegas, would be via the badge scans.
221 00:27:48.930 ⇒ 00:28:05.299 Katherine Bayless: with the, Unveiled Europe stuff, unfortunate same name, we do have that registration data in the organization, it’s just in Cvent, and we haven’t brought it into the data lake, but we can bring it in, it’s just not in there yet. So, like, right now.
222 00:28:05.300 ⇒ 00:28:13.819 Katherine Bayless: if you were to join with the CES, seed that Kyle’s put together, it would only really be the CES in Vegas that we’ve got the… yeah.
223 00:28:14.020 ⇒ 00:28:30.679 Kyle Wandel: Yeah, it was more for a Sweeney, because that’s… that has a lot of the… some of the unveiled exhibitor information. So, but I just added it all, just so we have a standardized table for all CES events, basically. So use, like, use that going forward. That’ll be a good bridge table.
224 00:28:31.970 ⇒ 00:28:38.439 Awaish Kumar: Okay, and, like, that… includes the dates for the events, or… Yep.
225 00:28:38.440 ⇒ 00:28:57.029 Kyle Wandel: the dates, the days, the start date, the end date, the location, and then any notes. Is it digital, because 2021’s digital? Is it primary? So is it the main show, or is it unveiled? So, yeah, there should be a good call-out for it. And I sent this through, I put it in Slack as well, but I can do it again.
226 00:28:57.690 ⇒ 00:29:07.539 Awaish Kumar: Okay, yeah, then we might convert it to being an event dim table, so it has information regarding the event, and joined with registration.
227 00:29:07.720 ⇒ 00:29:16.319 Awaish Kumar: And, yeah, for the attendant flag that Kyle was talking about.
228 00:29:16.550 ⇒ 00:29:25.269 Awaish Kumar: So, I thought, like, we should have, like, maybe a DMCES person table, which includes information regarding all the people.
229 00:29:25.400 ⇒ 00:29:31.229 Awaish Kumar: And, they are joined with, in fact, CS registration via email.
230 00:29:31.480 ⇒ 00:29:37.300 Awaish Kumar: And for that registration, And then we have a flag, right, if it is… if it…
231 00:29:37.630 ⇒ 00:29:40.749 Awaish Kumar: Was attended by the person or not, so…
232 00:29:41.480 ⇒ 00:29:49.920 Awaish Kumar: like, this is one of the ways. Other ways, like, suggested by Carl, to have a separate demo attendant. So what do you think we should be going with?
233 00:29:50.480 ⇒ 00:29:53.450 Katherine Bayless: Actually, I was,
234 00:29:53.450 ⇒ 00:30:18.379 Katherine Bayless: I realize, yeah, we should totally do it as dim CES person, Kyle, because then we can start with what we have now, registered, attended, canceled, we should put canceled in there too, where appropriate. But we could eventually also, we have plenty of people that we invite to CES that do not ever even, you know, touch the registration, but someday it might be nice to be able to also analyze
235 00:30:18.380 ⇒ 00:30:31.159 Katherine Bayless: like, CES person invited that never, you know, progressed past that stage. But if we start with registered, attended, and canceled, that’s good enough for the moment. But I like that it would be flexible to add more detail going forward.
236 00:30:31.710 ⇒ 00:30:40.969 Kyle Wandel: Yep, I agree. Just, like, make sure to… because I’m looking at it now, just make sure to add, like, the status. Did they… were they canceled, were they… were they just registered, or were they attended?
237 00:30:42.320 ⇒ 00:30:46.599 Awaish Kumar: Yeah, that, like, actually, that flag is infectable, basically.
238 00:30:46.830 ⇒ 00:30:47.610 Kyle Wandel: Okay.
239 00:30:47.610 ⇒ 00:30:51.690 Awaish Kumar: a distant ID, and there’s a connected flag if it was attended by
240 00:30:51.870 ⇒ 00:30:58.870 Awaish Kumar: by the person, or not, or… There’s also flag for is canceled, if registration was canceled.
241 00:30:59.240 ⇒ 00:31:02.639 Awaish Kumar: And then if we need to dig deeper into, like,
242 00:31:03.030 ⇒ 00:31:12.950 Awaish Kumar: the person information, the… that, like, the name, what is the name, or something like that, then we can join with it… join it with the MCS person table.
243 00:31:14.470 ⇒ 00:31:21.740 Katherine Bayless: Okay, okay, so then if we did want to expand it to include the invitee stuff, we would just need a different…
244 00:31:22.300 ⇒ 00:31:25.820 Katherine Bayless: fact table for, like, CES invites or something, I guess.
245 00:31:29.750 ⇒ 00:31:36.349 Awaish Kumar: Yeah, like, we can modify it. Right now, because it’s only the registration data, then it says vaccine registration.
246 00:31:36.600 ⇒ 00:31:42.430 Awaish Kumar: I can maybe rename it and include everything, and handle with the flags, whatever it is.
247 00:31:42.880 ⇒ 00:31:49.850 Katherine Bayless: Yeah, I mean, like I said, it doesn’t have to happen at this stage, it’s just kind of a future possibility, but, yeah.
248 00:31:55.940 ⇒ 00:31:59.139 Awaish Kumar: And then there’s, like, this bridge table.
249 00:31:59.470 ⇒ 00:32:13.059 Awaish Kumar: Which could be something where, if you are using these basic fact and dim tables, then it might become confusing how to join fact CSS registration, like the…
250 00:32:13.450 ⇒ 00:32:22.980 Awaish Kumar: like, the breakdown by product codes, or by product categories. We have a bridge because it’s a many-to-many relationship.
251 00:32:23.100 ⇒ 00:32:24.090 Katherine Bayless: Between.
252 00:32:24.090 ⇒ 00:32:29.759 Awaish Kumar: Fact registration and the… Product interest code, DIM table.
253 00:32:30.160 ⇒ 00:32:35.280 Awaish Kumar: So, to handle that, like, we came up with this, logic of bridging.
254 00:32:35.400 ⇒ 00:32:42.580 Awaish Kumar: So, this table will have the mappings between registration and product codes, and these three tables will
255 00:32:42.790 ⇒ 00:32:48.540 Awaish Kumar: Joined to basically answer the question regarding… The category.
256 00:32:49.590 ⇒ 00:32:50.970 Katherine Bayless: Okay, that makes sense.
257 00:32:51.100 ⇒ 00:33:00.400 Katherine Bayless: We could probably take the same approach for the show items, in that case, because that’s another sort of many-to-many, yeah.
258 00:33:00.930 ⇒ 00:33:03.400 Kyle Wandel: And it’s separated by, I think, semicolon.
259 00:33:03.980 ⇒ 00:33:05.009 Katherine Bayless: I think so, yeah.
260 00:33:09.200 ⇒ 00:33:21.320 Kyle Wandel: So yeah, there’s one… there’s one column called Show Item List, I believe it’s called that in the registration page, let me verify that, but essentially it’s, like, all of the codes that are associated with that registration.
261 00:33:21.880 ⇒ 00:33:24.149 Katherine Bayless: Yeah, like, all the things that they purchased, basically.
262 00:33:24.150 ⇒ 00:33:25.740 Kyle Wandel: Yeah, purchased, yeah, sorry.
263 00:33:26.670 ⇒ 00:33:42.539 Awaish Kumar: Okay, but is there any dimension for that? Like, for the product interest code, we have data and registration table, but then we also have a DEM table, a clean version of CS product interest code, where basically we’re joining codes with the names.
264 00:33:43.250 ⇒ 00:33:59.340 Katherine Bayless: Yeah, so in that CES filtration document, that would also have all of the show items and, like, what the codes are that are in our data, as well as what the full name is, and then also, like, what type of product, or item it was that they purchased, because we sell, like.
265 00:33:59.340 ⇒ 00:34:22.169 Katherine Bayless: registration, obviously, everybody buys that. And then we also sell, like, session tracks, so, like, additional packets of, or, sets of conference content, and then we also sell, like, reports, and things like that, and so we would be able to create a dimension around the show items, and we would also be able to track, like, each year what was offered, that kind of thing.
266 00:34:22.170 ⇒ 00:34:29.599 Katherine Bayless: I think right now, that master filtration draft is just the past years, but we could eventually add previous to it.
267 00:34:30.850 ⇒ 00:34:33.480 Awaish Kumar: Okay, and, like, what is it called? Master?
268 00:34:33.790 ⇒ 00:34:36.579 Katherine Bayless: It’s called, Show Items List.
269 00:34:36.580 ⇒ 00:34:40.949 Kyle Wandel: I’m looking to see if I named that something else in the staging, real quick.
270 00:34:42.110 ⇒ 00:34:47.940 Kyle Wandel: Nope, it’s still show item list. In the… base registration, it should be.
271 00:34:48.889 ⇒ 00:34:49.549 Awaish Kumar: Okay.
272 00:34:50.079 ⇒ 00:34:57.349 Awaish Kumar: Yeah, I can bring that in, and… No.
273 00:34:57.950 ⇒ 00:35:22.830 Katherine Bayless: Eventually, we’ll be able to create more of, like, a true sort of CES ledger, because I do know from the team over at Merits that the show item list that we get, like, because we get, like, one row per person, we get all of the show items, and then we get a price that they paid, and we don’t necessarily know, like, whether they paid full price for each of those items, or, you know, if, like, one was free, but normally is charged, but then they paid for something
274 00:35:22.830 ⇒ 00:35:25.299 Katherine Bayless: else, and now that data’s kind of lost, and so…
275 00:35:25.300 ⇒ 00:35:28.050 Katherine Bayless: Getting the, in fact, or the actual, like.
276 00:35:28.300 ⇒ 00:35:36.779 Katherine Bayless: amount paid for the items, we need to get more data from Merits eventually. But at least starting with what they bought is a, you know.
277 00:35:36.960 ⇒ 00:35:38.129 Katherine Bayless: Better than nothing.
278 00:35:44.010 ⇒ 00:35:47.149 Awaish Kumar: Okay, yeah, I think that’s… that’s it on the modeling.
279 00:35:48.640 ⇒ 00:36:01.379 Katherine Bayless: Cool. This is great. I think I’ll take some more time this afternoon and go through, now that I have a little bit better of an understanding of kind of how this is all meant to function, if you can share that doc in Slack, or if it’s already there, I go find it.
280 00:36:01.380 ⇒ 00:36:04.259 Awaish Kumar: A little bit, add more information and send it in on the slide.
281 00:36:04.630 ⇒ 00:36:05.620 Katherine Bayless: Okay, cool.
282 00:36:05.790 ⇒ 00:36:08.239 Katherine Bayless: Then I guess, Utam, if you’d like to…
283 00:36:08.240 ⇒ 00:36:09.180 Uttam Kumaran: Yes.
284 00:36:09.510 ⇒ 00:36:10.600 Katherine Bayless: on AI things.
285 00:36:10.880 ⇒ 00:36:18.869 Uttam Kumaran: Yeah, so let me just push… I’m just… was trying to fix this one thing. Let me just push all this to a PR, and I’ll just walk through some documents, and then I can share a demo.
286 00:36:22.590 ⇒ 00:36:25.229 Katherine Bayless: If you need me to buy you a few minutes of time, I can…
287 00:36:25.230 ⇒ 00:36:27.739 Uttam Kumaran: Yeah, maybe, like, in 90 seconds.
288 00:36:27.740 ⇒ 00:36:45.999 Katherine Bayless: Okay, yeah, yeah, so real quick, Fortune 500 stuff. So Kyle and I had a good chat about this, yesterday, and I did put, just before this call into the data lake, I created a directory for seeds, and then Fortune 500, and then CES year 2026 was just the first, sort of, seed file.
289 00:36:46.000 ⇒ 00:37:03.529 Katherine Bayless: But because it’s really important that once we tell the media how many Fortune 500 companies were at CES, we always then give the same answer, we’re gonna… we’ll create the seeds for the previous years, but basically, we’ll have this be structured at the full Fortune 500 list.
290 00:37:03.530 ⇒ 00:37:26.329 Katherine Bayless: So the rank, the list year, the company name on the Fortune 500 list, and then just a Boolean for, like, were they at CES, true or false? And so, if we had a question around, like, you know, how many people from Fortune 500 companies came, then we would go through the identity stitching to get to that registration data, but no longer storing the Fortune 500 flag with
291 00:37:26.330 ⇒ 00:37:35.200 Katherine Bayless: the attendee and registration data, where it’s just the companies are too messy. So this way, we’ll have a canonical, correct answer for how many Fortune 500 companies came.
292 00:37:38.230 ⇒ 00:37:42.889 Awaish Kumar: So, like, we already have one table for Fortune 500,
293 00:37:43.030 ⇒ 00:37:45.499 Awaish Kumar: Do we want to use that, or do we know…
294 00:37:45.990 ⇒ 00:37:58.810 Katherine Bayless: I think we wanted to make some changes to that as it was structured currently, or at least update it to draw from the new seed files. I can’t remember exactly what the structure was for it. Let me take a look.
295 00:38:02.340 ⇒ 00:38:06.280 Awaish Kumar: Yeah, we can, like, just take the same table and… and…
296 00:38:07.730 ⇒ 00:38:14.090 Awaish Kumar: added in that CS modeling as a DIM Fortune 500 companies, So that it’s there.
297 00:38:14.410 ⇒ 00:38:20.980 Katherine Bayless: Yeah, yeah, sorry, yeah, yeah, so you had the Fortune 500 file that was in the raw archive data?
298 00:38:21.830 ⇒ 00:38:33.820 Katherine Bayless: Yeah, so that’s basically… we’re gonna use that file to generate the prior years, but then they’ll all look like the seed file that I put in S3. But yeah, it’ll essentially be the same data, we’re just gonna structure it
299 00:38:34.900 ⇒ 00:38:35.880 Katherine Bayless: Differently.
300 00:38:45.150 ⇒ 00:38:46.399 Uttam Kumaran: Okay, I’m ready.
301 00:38:46.580 ⇒ 00:38:47.769 Katherine Bayless: Okay, let’s go.
302 00:38:47.770 ⇒ 00:38:51.220 Uttam Kumaran: Alright, so I’m just gonna walk through,
303 00:38:52.070 ⇒ 00:39:05.170 Uttam Kumaran: I’m just gonna walk through, like, what we were able to do, and then, it’s been interesting learning, like, about all the various pieces that are open, and then I’ll kind of share a demo, and then I think we can talk about,
304 00:39:05.400 ⇒ 00:39:11.520 Uttam Kumaran: Security, and, like, kind of, like, how we want to start to open this up, and, like, really, hopefully, this team can do a lot of testing this week.
305 00:39:11.740 ⇒ 00:39:30.600 Uttam Kumaran: But a couple things. So, I wrote this, like, little README, which just helps, like, how we have stuff set up. I focus on just badge skin data as just, like, one clear, like, area to focus on. There are quite a bit of,
306 00:39:30.920 ⇒ 00:39:50.150 Uttam Kumaran: like, objects to create to, like, enable really rich semantic context, so I was like, let me just go deep on, like, one piece, and sort of do all the pieces. And so, I think… I’ve also committed to the… to… I think it’s in this README, like, some helpful Medium articles and docs.
307 00:39:50.390 ⇒ 00:40:04.770 Uttam Kumaran: Just, like, in case you want to do further reading, but, basically, like, a semantic views is… is kind of exactly what it means. It’s just a view of, like, all of the semantic definitions, dimensions, metrics, relationships.
308 00:40:04.930 ⇒ 00:40:21.610 Uttam Kumaran: And so this is where we’re gonna be able to put in all the, like, rich information, about, like, our tables and the way we want people to answer. And so as part of this, I think we previously talked about several concepts around how do we continue to shape
309 00:40:21.780 ⇒ 00:40:40.999 Uttam Kumaran: our semantic layer in Snowflake, to allow for people to answer these questions. And so, I was able to build one… I just picked, like, one source table, and this is all in my dev mart right now. I created one, like, semantic view, and then I added comments and descriptions, and so…
310 00:40:41.210 ⇒ 00:40:50.739 Uttam Kumaran: one, I think, I can walk through, like, how we created this, like, semantic view table. So I think this is,
311 00:40:51.830 ⇒ 00:40:56.419 Uttam Kumaran: Yeah, let me just show you, like, an example. So, this is a semantic view,
312 00:40:56.680 ⇒ 00:41:04.969 Uttam Kumaran: I think, like, net-net, you have… it looks kind of similar to a great table statement, but you have comments.
313 00:41:05.220 ⇒ 00:41:19.660 Uttam Kumaran: You have, some, like, metrics here. So these are, like, common aggregations or, like, common dimensions that are being used. It’s sort of interesting, like, I, I didn’t,
314 00:41:20.930 ⇒ 00:41:35.410 Uttam Kumaran: it’s sort of kind of similar to what you would do in, like, a Looker, or, like, an Omni, where you… or, you know, where you, like, select what is a dimension and what is a metric, and then the metric, you have the aggregation. It’s interesting enough, it’s, like.
315 00:41:35.540 ⇒ 00:41:39.169 Uttam Kumaran: I think they… they…
316 00:41:39.950 ⇒ 00:41:58.619 Uttam Kumaran: I’ll kind of walk us through, like, even further than these types of objects, but you can start to label, like, ways you want things to commonly be aggregated if, like, a certain question is asked, and then we can add comments. And so one thing that, like, we can do is just really stuff this full of, like.
317 00:41:58.700 ⇒ 00:42:11.269 Uttam Kumaran: really, really great stuff. But I’ve gone and created semantic views for, a few, tables. And then the next piece is gonna be, yeah, go ahead.
318 00:42:11.790 ⇒ 00:42:28.900 Katherine Bayless: So, I think if I follow what you’re talking about under metrics for the badge scans, for example, like, we could have, you know, the metric of total scans is count star, the metric of, like, total people scanned would be count distinct registrant ID. Like, we would put that definition in there.
319 00:42:29.860 ⇒ 00:42:31.020 Uttam Kumaran: Yes, that’s correct.
320 00:42:31.450 ⇒ 00:42:32.490 Katherine Bayless: Cool.
321 00:42:32.490 ⇒ 00:42:46.990 Uttam Kumaran: Correct. And so you’ll also see that I’ve, I’ve added what, I’ve added, like, canonical queries, so these are helpful for us to, like, validate the Cortex analysts. So these aren’t, like, Snowflake
322 00:42:47.060 ⇒ 00:43:04.670 Uttam Kumaran: specific things, but for us, as always what I mention, is, like, we want to be able to test that the way, like, we… the way we ask a question gets answered in this way. And so I wanted to go ahead and, like, seed our repo with, like, what are some, like.
323 00:43:04.960 ⇒ 00:43:17.000 Uttam Kumaran: we would call it, like, the, like, golden data set. Like, what are the queries that are the source of truth answers for this? And kind of as I wrote here, we’re using them to validate Cortex.
324 00:43:17.230 ⇒ 00:43:34.620 Uttam Kumaran: we’re using them to make sure that the logic is exactly the same, and then when we do demos, it’s actually really helpful to know, like, what are some, like, safe things that demo really well. So I think that’s, like, kind of, like, why… this is not Snowflake, specific, but…
325 00:43:34.800 ⇒ 00:43:46.150 Uttam Kumaran: when I think about not only ensuring that we’re confident in the answers, but that when we go to share people, we have some things to try that demonstrate the power, it’s… we have some…
326 00:43:46.290 ⇒ 00:43:51.340 Uttam Kumaran: you know, evidence for that. So let me just, like… Close some of these.
327 00:43:51.850 ⇒ 00:43:59.210 Uttam Kumaran: So… That’s that. I think the next piece is…
328 00:43:59.390 ⇒ 00:44:03.770 Uttam Kumaran: these are some of the ways I thought about, like, easy, medium, hard.
329 00:44:04.120 ⇒ 00:44:10.010 Uttam Kumaran: questions. I think this is, again, really helpful for us to…
330 00:44:10.450 ⇒ 00:44:23.370 Uttam Kumaran: I just took a first pass at these, but we should work on these as a group. But I am curious if, like, any of these make sense, or if there are any at the top of your head, especially for, like, I’m more concerned about medium and hard.
331 00:44:23.690 ⇒ 00:44:30.940 Uttam Kumaran: where, like, yes, there are complicated joins, or, like, window functions, or aggregates, or, like.
332 00:44:31.270 ⇒ 00:44:36.570 Uttam Kumaran: only, like, CTA-specific lingo… Yeah. Yeah.
333 00:44:36.570 ⇒ 00:44:46.900 Katherine Bayless: I think a great hard one for the badge skins in particular is one that Kyle’s working on with, Jackie, the VP of our conference’s team, which is, like.
334 00:44:47.040 ⇒ 00:44:55.450 Katherine Bayless: For people who purchased a track, so that show item list, that gave them access to certain sessions, did they go?
335 00:44:55.780 ⇒ 00:44:59.040 Katherine Bayless: And so it’s sort of like an attendance metric.
336 00:44:59.040 ⇒ 00:44:59.660 Uttam Kumaran: Yeah.
337 00:44:59.660 ⇒ 00:45:06.110 Katherine Bayless: be, like, how many people bought a track and actually attended the things that they paid extra for.
338 00:45:06.230 ⇒ 00:45:17.429 Katherine Bayless: part of why we’re doing that is the obvious analysis, but then also there’s a… I guess somewhere along the lines we learned that, like, people tend to buy a track one year and then not the next year.
339 00:45:17.430 ⇒ 00:45:17.790 Uttam Kumaran: Okay.
340 00:45:17.860 ⇒ 00:45:24.160 Katherine Bayless: So, we’re kind of curious to get a sense of, like, you know, what are the behaviors really around these track purchases?
341 00:45:24.690 ⇒ 00:45:28.199 Uttam Kumaran: Okay, I just typed that in, and yeah, I’m just gonna… I’ll add that to…
342 00:45:28.500 ⇒ 00:45:39.080 Uttam Kumaran: this hard section, but that’s great. So again, I think it just gives us, like, a sense, and this… this may be just for… this may be broader, this may be just per team, but…
343 00:45:39.270 ⇒ 00:45:45.720 Uttam Kumaran: this is sort of… I just wanted to set, like, the foundation of, like, how or where we kind of think about orchestrating this.
344 00:45:47.120 ⇒ 00:45:47.880 Katherine Bayless: Yeah.
345 00:45:47.880 ⇒ 00:45:48.879 Uttam Kumaran: Is there any other… yeah.
346 00:45:49.030 ⇒ 00:45:59.309 Katherine Bayless: a lot of work on kind of identifying some of these metrics, too, so this is a great way to kind of take that, take all of that stuff that Kai’s pulled together and put it into this sort of
347 00:45:59.640 ⇒ 00:46:00.690 Katherine Bayless: structure.
348 00:46:01.470 ⇒ 00:46:15.620 Chi Quinn: Yeah, because I was going to ask, because I did provide, just, like, the first set of metrics, and I’m curious to know, like, what other content or information should I put in that would satisfy the…
349 00:46:15.920 ⇒ 00:46:18.740 Chi Quinn: the needs for cortex.
350 00:46:18.950 ⇒ 00:46:22.800 Uttam Kumaran: Yeah, one thing that’s helpful, even, and we could do it as part of this, is, like.
351 00:46:22.800 ⇒ 00:46:23.890 Chi Quinn: What was the…
352 00:46:23.920 ⇒ 00:46:29.269 Uttam Kumaran: previous Power BI that, like, was supporting this question, potentially.
353 00:46:29.320 ⇒ 00:46:33.360 Chi Quinn: That way, it allows this to be more real. Like, I was just kind of…
354 00:46:33.360 ⇒ 00:46:39.469 Uttam Kumaran: Brainstorming, but allows us to maybe be tied directly to past queries that we know the business asks.
355 00:46:39.740 ⇒ 00:46:51.340 Uttam Kumaran: Of course, like, I… I’m just pulling it from, sort of, our conversations, but that way it allows us to say, like, well, you were using this dashboard, and here’s, like, all the questions that that dashboard answered.
356 00:46:51.590 ⇒ 00:46:52.630 Chi Quinn: Like.
357 00:46:53.000 ⇒ 00:46:53.560 Katherine Bayless: Yeah.
358 00:46:53.560 ⇒ 00:46:59.160 Uttam Kumaran: like, and here’s how they can now be answered here. It’s sort of like, really trying to make a one-to-one.
359 00:47:00.120 ⇒ 00:47:00.500 Uttam Kumaran: You know…
360 00:47:01.290 ⇒ 00:47:02.300 Katherine Bayless: Very cool.
361 00:47:02.980 ⇒ 00:47:03.710 Uttam Kumaran: Yeah.
362 00:47:04.780 ⇒ 00:47:07.660 Katherine Bayless: I do think we’ll probably eventually have them by team, yeah, because I can also.
363 00:47:07.660 ⇒ 00:47:08.430 Uttam Kumaran: Yeah.
364 00:47:08.430 ⇒ 00:47:15.710 Katherine Bayless: start creating these for, like, the membership folks, right? Like, easy questions don’t seem to exist. They’d all be hard.
365 00:47:15.710 ⇒ 00:47:25.889 Uttam Kumaran: Yeah, and also, you can think about, like, the kind of unit test model is, like, I want to be able to run our questions and enforce that the answers are being
366 00:47:26.510 ⇒ 00:47:30.859 Uttam Kumaran: The answers are being shared in the way we expect?
367 00:47:30.970 ⇒ 00:47:39.010 Uttam Kumaran: Like, all the time, especially as we make changes. So, some of these we will know the answer to, and we’ll hard code it, and that’ll be, like, the unit test
368 00:47:39.210 ⇒ 00:47:51.190 Uttam Kumaran: Right? Previously, unit tests, it was very similar, but, like, it was actually less about, using AI to ask a question, it was, like, run the SQL query if you get this answer, we’re, like, still in good shape, you know?
369 00:47:51.520 ⇒ 00:47:52.190 Katherine Bayless: Yeah.
370 00:47:53.660 ⇒ 00:48:00.120 Uttam Kumaran: Okay, cool. So, yeah, there’s a little bit of… yeah, we ran this,
371 00:48:00.980 ⇒ 00:48:07.270 Uttam Kumaran: blah blah blah blah blah… Okay, and then I’ll come back and talk about security, because that’s what I was, like.
372 00:48:07.890 ⇒ 00:48:10.750 Uttam Kumaran: messing around with right before this.
373 00:48:13.060 ⇒ 00:48:16.730 Uttam Kumaran: So, I… maybe I’ll kind of just walk through, like, from, like, a…
374 00:48:16.880 ⇒ 00:48:33.360 Uttam Kumaran: I mean, I guess it’s… it would just be me closing this, but basically what you could do is you could just click on this thing on the right if you haven’t already, and I’ll go ahead and just, like, click Create a New Cortex code. And you don’t necessarily, like, need to have this open, but…
375 00:48:33.420 ⇒ 00:48:42.439 Uttam Kumaran: you know, like, I think it… if relevant to your question, like, if you’re a developer, and you’re writing some queries, and then you want to switch between Cortex code.
376 00:48:42.580 ⇒ 00:48:52.039 Uttam Kumaran: it’s nice to do that, but, like, for example, I just want to say, like, tell me about our badge scan data,
377 00:48:52.220 ⇒ 00:49:06.139 Uttam Kumaran: the one, like, there’s gonna be a little bit of art on, like, how to write these questions. Second, like, there’s things around skills and things like that. I haven’t really gone too far into that world. And then there’s, like, they have some out-of-the-box models.
378 00:49:06.330 ⇒ 00:49:09.939 Uttam Kumaran: for our work, I feel like…
379 00:49:10.600 ⇒ 00:49:28.280 Uttam Kumaran: it’s hitting auto is probably fine. I think we’ll also probably end up seeing, like, how does the billing look like on these, and, like, limiting. That’s what I’m actually more concerned about, but, like, just, like, keep it simple right now. Let’s say, tell me about our badge scan data.
380 00:49:28.400 ⇒ 00:49:33.479 Uttam Kumaran: What you’re gonna see it do is… it’s gonna share you a little bit of its thinking.
381 00:49:33.670 ⇒ 00:49:46.509 Uttam Kumaran: So again, if you’ve used Kersher or Copilot, you’ll rec… or ChatGP, you’ll recognize this. And then it’s gonna look through and look at this. So one thing that I can do right now is I could actually… I’m just using my,
382 00:49:46.980 ⇒ 00:49:49.399 Uttam Kumaran: like, I’m just using my dev,
383 00:49:50.160 ⇒ 00:49:55.030 Uttam Kumaran: my dev schema, so I can say,
384 00:49:55.220 ⇒ 00:50:00.890 Uttam Kumaran: Tell me about the categories of scan data we have.
385 00:50:16.430 ⇒ 00:50:24.840 Uttam Kumaran: And then… 70 points… It’s able to sort of… Answer.
386 00:50:25.260 ⇒ 00:50:31.570 Uttam Kumaran: And so, one thing that is nice about this is not only do you see the evidence of, like, what it’s…
387 00:50:31.700 ⇒ 00:50:32.969 Uttam Kumaran: Trying to do.
388 00:50:33.160 ⇒ 00:50:39.760 Uttam Kumaran: And so some things right now, it’s still going into raw. So one of the things I’m trying to figure out is, like, how do we steer
389 00:50:40.300 ⇒ 00:50:46.439 Uttam Kumaran: because I think my role has access to everything, it’s gonna look through everything, so how do we steer it to just…
390 00:50:46.700 ⇒ 00:50:48.649 Uttam Kumaran: Pull from the things we create.
391 00:50:48.770 ⇒ 00:50:50.360 Uttam Kumaran: But…
392 00:50:50.360 ⇒ 00:51:04.230 Katherine Bayless: Like, I’ve been often saying, like, using just the data in either, like, this database, this schema, or even, like, you know, this table or view specifically, but yeah, like, most users aren’t gonna know that or want to do that, but .
393 00:51:04.620 ⇒ 00:51:14.040 Uttam Kumaran: Yeah, that was sort of the testing that I was doing, like, maybe I can even… I can even have it do that, like, using just the information here.
394 00:51:16.010 ⇒ 00:51:35.099 Uttam Kumaran: But I wanted to kind of give you a little bit of UX of, like, this is how people are going to get displayed, and so I still… I’m not 100% sure how much control we have over this, but I know we can… we can steer a lot of how things are getting responded to. So there’ll be, like, a table output, and then it should give some helpful things, and you’ll actually see the evidence here.
395 00:51:35.360 ⇒ 00:51:38.410 Uttam Kumaran: Which is really nice. And then…
396 00:51:38.730 ⇒ 00:51:42.060 Uttam Kumaran: Yeah, so now this is, like, what I… what I kind of…
397 00:51:42.260 ⇒ 00:51:45.809 Uttam Kumaran: Wanted to see, which is basically,
398 00:51:46.050 ⇒ 00:51:52.009 Uttam Kumaran: It’s gonna go look through… it’s gonna run a describe view, which is gonna give it all of that semantic context.
399 00:51:52.220 ⇒ 00:51:57.000 Uttam Kumaran: It’s gonna then… I shoot the query, which is, like, get the query.
400 00:51:57.390 ⇒ 00:52:05.059 Uttam Kumaran: As you can see, there are some great descriptions, that I wrote, and then it’s going to, pull those out.
401 00:52:05.280 ⇒ 00:52:13.119 Uttam Kumaran: So, like, that… that… this is… I would say this is, like, the first part of, like, just…
402 00:52:13.510 ⇒ 00:52:14.490 Uttam Kumaran: the…
403 00:52:14.700 ⇒ 00:52:30.480 Uttam Kumaran: what is the… how do you set up each of the pieces, of, like, the semantic layer? Like, how does… how do we basically enforce this? And then I have some pieces that I can talk about, like, a little bit about security, and the security piece, actually.
404 00:52:30.600 ⇒ 00:52:39.269 Uttam Kumaran: Is going to affect this kind of problem that we saw, which is, like, scoping a role to just certain things, so that it doesn’t go…
405 00:52:39.920 ⇒ 00:52:44.450 Uttam Kumaran: Like, it’s… what it probably did is it literally, like, did a… it did a…
406 00:52:44.770 ⇒ 00:52:52.190 Uttam Kumaran: table scan for anything that was, like, had badge or something in it, and, like, went straight to there, you know? So, any questions, like…
407 00:52:52.740 ⇒ 00:52:53.920 Uttam Kumaran: So far.
408 00:52:54.290 ⇒ 00:52:58.769 Uttam Kumaran: I mean, this is… I feel like we’re all data people, this isn’t, like, the craziest thing, but…
409 00:52:58.950 ⇒ 00:53:00.350 Uttam Kumaran: I don’t know, it’s .
410 00:53:01.180 ⇒ 00:53:14.030 Katherine Bayless: No, I mean, I think, honestly, like, at least for me, like, I’ve been using Coco, really extensively, so getting a sense of what we can do to help it, like, do better answering the ques… I mean, it does really great without.
411 00:53:14.030 ⇒ 00:53:15.450 Uttam Kumaran: Yes. Yes.
412 00:53:15.450 ⇒ 00:53:29.909 Katherine Bayless: the already great responses, is a really kind of exciting place for us to start playing. I have some thoughts about things we could add, but, Kyle or Kai, if you’ve got thoughts for…
413 00:53:30.330 ⇒ 00:53:41.550 Chi Quinn: Yeah, I was thinking… I don’t know, just on top of my head, like, some type of little summary, because I’m thinking from the end user, when they… when the number or the answer they give spits out, they might want to
414 00:53:41.550 ⇒ 00:53:51.270 Chi Quinn: further explanation, or something that just emphasized that, oh, this came from this area, whatever the language that’s, I guess, the language
415 00:53:51.380 ⇒ 00:53:53.790 Chi Quinn: Just in plain English on the bottom.
416 00:53:53.790 ⇒ 00:54:00.479 Uttam Kumaran: Can you give me a… can you… can… yeah, I think one thing that’d be helpful for me is, like, a good… if we would use an example, you don’t have to use batch scans, but, like…
417 00:54:00.640 ⇒ 00:54:05.840 Uttam Kumaran: I think it’s narrated out, like, what you would expect to see given, like, an input question.
418 00:54:06.100 ⇒ 00:54:07.320 Katherine Bayless: Yeah.
419 00:54:12.010 ⇒ 00:54:16.200 Chi Quinn: I don’t know, I was… yeah, oh, go ahead, Kyle. You might know way more.
420 00:54:16.200 ⇒ 00:54:32.099 Kyle Wandel: Yeah, I was just gonna say, from my perspective, I think it looks pretty good. I think it’s okay if, I mean, it does an okay job of answering the questions that they want, but I wonder how it handles more in-depth analysis. I use it here and there, but, the simple stuff it already does a really good job with, for sure.
421 00:54:35.730 ⇒ 00:54:36.859 Uttam Kumaran: Yeah, I think…
422 00:54:37.110 ⇒ 00:54:44.789 Uttam Kumaran: I’m still in… yeah, I think… I agree, I think it’s doing good with the simple stuff. I think this is where we’re gonna start to test the more complicated questions.
423 00:54:44.980 ⇒ 00:54:46.360 Katherine Bayless: I still think, like.
424 00:54:46.400 ⇒ 00:54:54.350 Uttam Kumaran: Kai, I would love to hear about, like, what is the output, like, that example, because that’ll help me craft a system prompt for this.
425 00:54:54.610 ⇒ 00:54:55.410 Uttam Kumaran: You know?
426 00:54:55.410 ⇒ 00:54:55.750 Chi Quinn: Yeah.
427 00:54:59.110 ⇒ 00:54:59.560 Katherine Bayless: Yeah, like…
428 00:55:00.020 ⇒ 00:55:02.660 Chi Quinn: Yeah. Well, no, I was gonna say, like, just…
429 00:55:03.100 ⇒ 00:55:16.159 Chi Quinn: I’m thinking from attendee. I was thinking CES attendees, for example, like, someone might enter a specific question, like, who are… how many senior level attendees attended CES?
430 00:55:16.170 ⇒ 00:55:28.329 Chi Quinn: And usually, or at least what I saw, there was, some conflict on how we define senior-level attendees. So some might include board members, and something,
431 00:55:28.330 ⇒ 00:55:39.509 Chi Quinn: telco could spit out the answer, like, oh, blank 500 senior-level attendees attended, and then I guess the bottom part would say, this includes
432 00:55:39.510 ⇒ 00:55:42.089 Chi Quinn: People who are C-level, president.
433 00:55:42.090 ⇒ 00:55:43.260 Uttam Kumaran: Okay, okay.
434 00:55:44.110 ⇒ 00:55:57.829 Kyle Wandel: That makes sense, because I… yeah, I agree with Kai. They’re not gonna… they’re more than likely not going to click the queries that you saw, and, like, we understand that, but they may not do that. They may… some of them may, some of them might, but it might just be easier to just say.
435 00:55:57.830 ⇒ 00:56:03.060 Uttam Kumaran: So almost having, like, having some, like, here’s how I got here in, like, English.
436 00:56:03.530 ⇒ 00:56:11.090 Kyle Wandel: Yeah, basically. Like, it’s almost like… I wonder if Cortec can literally just pick it up as simple as, explain the where logic in English.
437 00:56:11.530 ⇒ 00:56:12.170 Uttam Kumaran: Yeah.
438 00:56:12.370 ⇒ 00:56:13.140 Uttam Kumaran: Okay.
439 00:56:14.610 ⇒ 00:56:15.859 Uttam Kumaran: Okay, that’s helpful.
440 00:56:17.960 ⇒ 00:56:40.300 Katherine Bayless: I think, too, like, so two additional places, because, yeah, I love the suggestion around, like, you know, explain what you did, right? I think the other thing might be interesting… some of it, I think, would be handled in, like, the defining the default aggregations and stuff, but, like, maybe, like, encouraging Coco to give little nudges to the user, like.
441 00:56:40.300 ⇒ 00:56:54.930 Katherine Bayless: So an example of one early adopter scenario we ran into was Dave Hennessy on our marketing team asked it for the top 20 countries at CES, and it gave him the right answer, but it didn’t rank the list in order.
442 00:56:54.930 ⇒ 00:56:55.730 Uttam Kumaran: Yeah.
443 00:56:55.730 ⇒ 00:57:05.080 Katherine Bayless: And so, I mean, you know, no big deal, fixed it, it’s fine. But, like, if Coco could have said, you know, like, would you like these ranked, or maybe we just tell it if somebody asks top X.
444 00:57:05.080 ⇒ 00:57:13.489 Uttam Kumaran: Always rank, yeah. Kind of a couple of things I’m always thinking about is, like, what is our standards, and, like, how can we implement our standards? And then second.
445 00:57:13.630 ⇒ 00:57:22.699 Uttam Kumaran: it should give, like, I think the best thing that I love when I use AI tools is when it says, like, here’s ways I can go further. Alright, and they kind of did, like, one of that.
446 00:57:22.840 ⇒ 00:57:32.800 Uttam Kumaran: But, like, I would like… we should program it and give our sense of, like, it should not only say, here’s, like, the way I got to this, and, like, the assumptions I made.
447 00:57:33.060 ⇒ 00:57:41.449 Uttam Kumaran: like, are they right? And then, here are ways that we can go one step further, because I think people are gonna find that those suggestions are helpful.
448 00:57:41.510 ⇒ 00:57:43.269 Katherine Bayless: And then, like, keep…
449 00:57:43.270 ⇒ 00:57:48.869 Uttam Kumaran: Like, asking those, and you want it to be more of, like, a brainstorming partner on, like, where could we go further?
450 00:57:49.720 ⇒ 00:58:13.800 Katherine Bayless: Yeah, exactly, and along those lines, too, when I’ve been working with some of the remembers data for the Journey stuff that we’re putting together, like, there have been instances where I’m getting counts that, like, I know just don’t feel quite right, and so I’ve actually gone into the, like, UI side and grabbed, like, an example of a record, and gone back to Coco and said, like, this record should appear in the dataset as X, Y, or Z.
451 00:58:13.800 ⇒ 00:58:15.760 Katherine Bayless: Can you use it for testing?
452 00:58:15.760 ⇒ 00:58:20.130 Katherine Bayless: And so, like, if there’s a way for it to suggest
453 00:58:20.360 ⇒ 00:58:28.150 Katherine Bayless: the user go get some test records in cases where a, like, metric is disputed or ambiguous. That could be kind of interesting.
454 00:58:28.150 ⇒ 00:58:28.930 Uttam Kumaran: Yeah.
455 00:58:29.820 ⇒ 00:58:38.100 Uttam Kumaran: Yeah, that’s… I’m also wondering, now that you said that, about, like, how this can take advantage of, or have knowledge of a lot of our DPT joins.
456 00:58:38.680 ⇒ 00:58:43.410 Uttam Kumaran: We are gonna indicate, like, how things work, but I’m… that’s something I’m gonna write down.
457 00:58:45.340 ⇒ 00:58:49.700 Awaish Kumar: Yeah, maybe, like, the queries we have created in Semantic Clear.
458 00:58:49.850 ⇒ 00:58:55.799 Awaish Kumar: If we verify them, like, the numbers are fine, the carrier look fine, these could be example
459 00:58:56.120 ⇒ 00:58:59.099 Awaish Kumar: Can be added as a verified queries for the Cortex.
460 00:59:01.580 ⇒ 00:59:18.220 Katherine Bayless: It’s true. I think there are definitely some cases where we could do that. There are others where the examples would change too often to want to park them in the example queries. Like, if you’re trying to get, like, a correct number for companies that are under account management light.
461 00:59:18.490 ⇒ 00:59:19.230 Uttam Kumaran: Yeah.
462 00:59:19.390 ⇒ 00:59:21.700 Katherine Bayless: sort of tag that’s assigned, and so you…
463 00:59:21.700 ⇒ 00:59:26.539 Uttam Kumaran: Must be, like, historical, like, a month, like, last year, like, something like that, yeah.
464 00:59:26.940 ⇒ 00:59:31.449 Katherine Bayless: We could totally do it for older, like, archived CES data and stuff like that.
465 00:59:32.750 ⇒ 00:59:40.270 Katherine Bayless: And then the other thought that I had was, knowing our users, and maybe this also sidles back Segway towards security,
466 00:59:40.560 ⇒ 00:59:44.030 Katherine Bayless: people are gonna want the export button, right?
467 00:59:44.030 ⇒ 00:59:44.410 Uttam Kumaran: Yeah.
468 00:59:44.700 ⇒ 00:59:45.710 Katherine Bayless: I think…
469 00:59:46.400 ⇒ 00:59:56.140 Katherine Bayless: a couple things. One is, knowing that we don’t have all the things in Snowflake right now, I am being more gentle on the export button amongst our colleagues, because, yeah, like.
470 00:59:56.140 ⇒ 00:59:56.590 Uttam Kumaran: Okay.
471 00:59:56.650 ⇒ 00:59:59.289 Katherine Bayless: Create too restrictive of an environment, and then they come out.
472 00:59:59.290 ⇒ 00:59:59.990 Uttam Kumaran: Yeah.
473 01:00:00.750 ⇒ 01:00:08.360 Katherine Bayless: But I would also like it if that export button was used sparingly, and it would be really cool if…
474 01:00:08.710 ⇒ 01:00:15.110 Katherine Bayless: we could have Coco be intelligent enough to, like, if somebody asks for the export button, like, ask them, like.
475 01:00:15.150 ⇒ 01:00:28.160 Katherine Bayless: can I redact the PII before I give you the export query? Like, that kind of thing. So, like, people might say, you know, I need a list of registrants at Fortune 500 companies, something like that.
476 01:00:28.160 ⇒ 01:00:39.589 Katherine Bayless: like, do you actually need their email addresses? Because if you don’t, let’s just not put them in that export. Let’s, you know, redact them with asterisks over the, you know, handle, or whatever sort of solution makes the most sense.
477 01:00:39.620 ⇒ 01:00:44.159 Katherine Bayless: But if there’s any way to incorporate some of that, like, you know… Yeah.
478 01:00:45.530 ⇒ 01:00:47.330 Katherine Bayless: I’m dreaming big, I know.
479 01:00:47.330 ⇒ 01:00:52.600 Uttam Kumaran: No, I mean, I think that’s helpful, like, I need… I want to start looking into how much we can turn on and off.
480 01:00:52.760 ⇒ 01:00:59.319 Katherine Bayless: And I think that’s gonna be some of the stuff I figure out with the security piece, so if I just can, like, go into that…
481 01:00:59.690 ⇒ 01:01:01.670 Uttam Kumaran: Briefly…
482 01:01:01.880 ⇒ 01:01:08.879 Uttam Kumaran: I was just playing around, but the demo was, like, kind of being wonky, but I’ll kind of share, like, what I was… what I was basically attempting to do.
483 01:01:09.010 ⇒ 01:01:12.900 Uttam Kumaran: Which is, like, have,
484 01:01:13.110 ⇒ 01:01:27.729 Uttam Kumaran: like, test out two roles, and then show that the Cortex was leveraging the role to query. Ultimately, in Snowflake World, everything runs through your role, so we’re not doing any grants at the user level.
485 01:01:27.790 ⇒ 01:01:42.550 Uttam Kumaran: Which is great, because for us, we can switch to the role and mimic people. For people, they, for the most part, they should just have the role that they’re tied to, like, one-to-one. They don’t really need to think much about, like, what do I have access to? Like, we would have thought about that.
486 01:01:42.590 ⇒ 01:02:00.430 Uttam Kumaran: And so one thing that I’m kind of thinking through is, like, an example of it working, where I wanted to show that you can run a count on badge scans with just, like, a normal role, and then with, like, the role, I have this, like, Cortex Limited role as a test.
487 01:02:00.470 ⇒ 01:02:08.789 Uttam Kumaran: And it should basically, like… it only had… it basically implemented, like, a where clause on the, on the table.
488 01:02:08.870 ⇒ 01:02:15.720 Uttam Kumaran: So the view, the way the view works is it’ll check your roll, and then… and then add that, and so…
489 01:02:16.200 ⇒ 01:02:33.999 Uttam Kumaran: that’s what I was sort of testing right before this, but wasn’t exactly working, but that’s sort of, like, what I wanted to try to show, is that, like, okay, it was actually, like, adhering to the right roles and the grants that were there. So I think another thing I can start to look at is what other…
490 01:02:34.110 ⇒ 01:02:48.169 Uttam Kumaran: Cortex, or even Snowflake broadly related, like, UI features we could limit, based on the role. Like, maybe people shouldn’t have works… maybe some people shouldn’t have worksheets, like, they only have Cortex.
491 01:02:49.990 ⇒ 01:02:54.130 Katherine Bayless: I mean, truthfully, that would be a good idea, more from, like, managing over.
492 01:02:54.130 ⇒ 01:02:56.140 Uttam Kumaran: Yes. Yes.
493 01:02:56.140 ⇒ 01:02:57.050 Katherine Bayless: Yeah, yeah.
494 01:02:57.050 ⇒ 01:03:03.670 Uttam Kumaran: Or, like, for me, I’m, like, thinking about, okay, should we accomplish that? Like, can I embed Cortex into the Streamlit app, then?
495 01:03:04.030 ⇒ 01:03:04.410 Katherine Bayless: And then…
496 01:03:04.410 ⇒ 01:03:18.850 Uttam Kumaran: like, so I’m kind of thinking… I’m doing some exploration more about, like, how it works. The other piece, is… I was just looking at this up before, is, like, there… if you guys noticed, thumbs up, thumbs down. So we do get feedback.
497 01:03:19.110 ⇒ 01:03:35.140 Uttam Kumaran: And so, that will sort of help for observability. So one thing I didn’t cover here is, like, how are we going to look at how many queries are coming, what people are asking, and the feedback? But I’ll… I’ll probably set up a quick dashboard in Snowflake for all of that, that we can start to look at, yeah.
498 01:03:35.920 ⇒ 01:03:50.070 Katherine Bayless: One question about the feedback mechanism, because I know, like, we… I’m pretty sure, at least, unless somebody’s turned it back on, we turned it off in, Claude, because when you, like, thumbs up the response, it sends that to Anthropic with the, like, you know, here was the question.
499 01:03:50.070 ⇒ 01:03:51.200 Uttam Kumaran: Oh…
500 01:03:51.370 ⇒ 01:03:56.439 Katherine Bayless: The user liked it, and we were like, yeah, that’s a great way to leak your data by accident, just because you wanted to tell the robot.
501 01:03:56.440 ⇒ 01:03:57.250 Uttam Kumaran: Yeah.
502 01:03:57.250 ⇒ 01:04:03.120 Katherine Bayless: And so I don’t know if Snowflake’s the same way, or if it is really just, like, local feedback that we’re collecting.
503 01:04:03.120 ⇒ 01:04:07.940 Uttam Kumaran: I can check, I mean, this is, like, sort of the doc I was reading about, like, the feedback.
504 01:04:09.080 ⇒ 01:04:09.639 Katherine Bayless: The…
505 01:04:09.640 ⇒ 01:04:14.370 Uttam Kumaran: like, the feedback on Cortex agents, and…
506 01:04:15.250 ⇒ 01:04:34.490 Uttam Kumaran: I don’t… I… I would be very, very surprised if, like, any of that is external. Like, I think that’s all our feedback. Like, none of it is regarding… because Snowflake has their own, like, help thing where you can submit feedback and things, so I don’t think this is related to that at all. Right. But I do see your point.
507 01:04:34.940 ⇒ 01:04:35.610 Katherine Bayless: Yeah.
508 01:04:38.670 ⇒ 01:04:44.199 Uttam Kumaran: Yeah, so that’s that. I think,
509 01:04:44.320 ⇒ 01:04:54.060 Uttam Kumaran: Yeah, on the dbt topic, I think, again, a lot of that we’re gonna try to just stuff into, like, the model and column descriptions, the logic.
510 01:04:54.390 ⇒ 01:04:58.429 Uttam Kumaran: And then, there is a,
511 01:04:59.590 ⇒ 01:05:09.850 Uttam Kumaran: like, there is a piece for… I was just, like, googling it right before when we were talking about, like, AI. Like, in the semantic view, you can actually add natural language rules.
512 01:05:10.170 ⇒ 01:05:13.310 Uttam Kumaran: So I think I’m gonna play around also with, like.
513 01:05:13.760 ⇒ 01:05:18.370 Uttam Kumaran: Showing how we can get the outputs to differ, and whether we can insert stuff, but…
514 01:05:18.530 ⇒ 01:05:31.500 Uttam Kumaran: there’s a lot of… there’s a lot of stuff they added that I’m, like, just walking through one by one to kind of learn about the pieces, but, I guess, like, in terms of next steps, I… I would love to see how we can drive towards…
515 01:05:33.190 ⇒ 01:05:40.670 Uttam Kumaran: like, a publicly… like, I think, and said another way, how can we start to grant some of our users an…
516 01:05:41.120 ⇒ 01:05:47.939 Uttam Kumaran: Like, basically, maybe have a meeting where we’re like, this is now available to you, here are the questions that you can…
517 01:05:48.120 ⇒ 01:05:53.180 Uttam Kumaran: ask, and here’s the types of underlying data it has access to.
518 01:05:53.560 ⇒ 01:06:00.429 Uttam Kumaran: And, like, work… basically, I’m like, when is that? And, like, how can we work backwards to now? And, like, I can start to think about how far we are from that.
519 01:06:01.130 ⇒ 01:06:08.329 Katherine Bayless: Yeah, so that is, in my mind, optimistically, March 30th.
520 01:06:08.490 ⇒ 01:06:09.160 Uttam Kumaran: Okay.
521 01:06:09.650 ⇒ 01:06:11.530 Katherine Bayless: Cool. Great.
522 01:06:11.530 ⇒ 01:06:13.490 Uttam Kumaran: Okay, helpful.
523 01:06:13.720 ⇒ 01:06:31.439 Katherine Bayless: Yeah, yeah, I mean, honestly, like, and I, you know, I know we will continue building, I know that there will be things that aren’t quite perfect out of the gate, but I think we’ll have the CES data in a good shape, if we have the views, at least around that, right? Like, I think we’ll be in a great place to, launch it to the organization on the 30th.
524 01:06:31.820 ⇒ 01:06:32.400 Katherine Bayless: Yeah.
525 01:06:32.400 ⇒ 01:06:33.440 Uttam Kumaran: So then I guess…
526 01:06:33.740 ⇒ 01:06:40.590 Katherine Bayless: I was gonna say, if the audit data doesn’t come back until April 1st, then we will hold, but yeah, March 30th is the date in my mind.
527 01:06:40.590 ⇒ 01:06:46.300 Uttam Kumaran: So then tell me, some other thoughts, like, for this team on, like, how we can, like.
528 01:06:46.510 ⇒ 01:06:51.890 Uttam Kumaran: do a demo, or whether we want to do, like, group training, and, like, how I can start to get organized around.
529 01:06:52.240 ⇒ 01:06:52.810 Katherine Bayless: Yeah.
530 01:06:53.010 ⇒ 01:07:04.909 Katherine Bayless: Yeah, I was gonna actually say we can talk about it kind of in-depth on planning on Monday, too, because I think we should kind of come up with a formal rollout plan to, like, how are we gonna… okay, you know, like, I need to write an email, and I will send an email.
531 01:07:04.910 ⇒ 01:07:05.390 Uttam Kumaran: Yeah.
532 01:07:05.390 ⇒ 01:07:18.359 Katherine Bayless: what are the… what are the things we want to offer people, to your point, like, office hours, workshops, one-on-one training, like, what do we want our suite of options to be? How, you know, do we want people to book these things? Because, like, if they want, you know.
533 01:07:18.360 ⇒ 01:07:18.990 Uttam Kumaran: Yes.
534 01:07:18.990 ⇒ 01:07:27.260 Katherine Bayless: you in the sessions and that kind of stuff, and so I think figuring out the logistics for rolling this out would be a good thing for us to all kind of put heads together on on Monday.
535 01:07:28.560 ⇒ 01:07:29.200 Uttam Kumaran: Yes.
536 01:07:30.870 ⇒ 01:07:34.550 Uttam Kumaran: So that would be great, if we can… yeah, if we can do that on Monday, that’d be perfect.
537 01:07:34.820 ⇒ 01:07:35.850 Katherine Bayless: Yeah.
538 01:07:36.930 ⇒ 01:07:54.299 Katherine Bayless: We’ll also, of course, as part of it, one of the, like, tiny logistics things is I’ll write up the, directions for people to, like, request Snowflake via Trellica, and so that way people know that they can ask for access to it and that kind of thing. And another homework item that’s right for me is to talk to Ian about
539 01:07:54.330 ⇒ 01:08:10.360 Katherine Bayless: right now, Trelica, when it provisions the user, it drops them in as, like, Snowflake Analyst, I think, and I am sure that is configurable, I just need to figure out how to have it drop them in in whatever role we want this sort of, early adopter Cortex-mainly, user to get.
540 01:08:12.690 ⇒ 01:08:13.650 Uttam Kumaran: Okay, okay.
541 01:08:14.880 ⇒ 01:08:16.120 Katherine Bayless: But yeah, this is exciting.
542 01:08:16.870 ⇒ 01:08:17.439 Uttam Kumaran: Perfect.
543 01:08:18.170 ⇒ 01:08:27.430 Katherine Bayless: Oh, another thought, too, for the semantic stuff, and maybe, Kai, you’ve probably already got some of this pulled together, we could work on, the missing pieces, but, like.
544 01:08:27.560 ⇒ 01:08:38.849 Katherine Bayless: for the different, sort of, like, data themes, if you will. So, like, badge scans, for example, including in the semantic view, like, this data comes from, you know.
545 01:08:38.850 ⇒ 01:08:58.739 Katherine Bayless: room entry scans, those devices are provided by Merits, the internal point of contact for questions on this data is, kind of a thing. And so that way, like, people, if applicable, there would be some of that information around, like, you know, well, where did we get this data, and who do I talk to if I have questions? Like, I think that would be helpful, too.
546 01:09:00.430 ⇒ 01:09:07.720 Kyle Wandel: I concur. I don’t want to be the person that gets 10,000. Where did this data come from? I don’t know how to go ask that person.
547 01:09:07.729 ⇒ 01:09:08.869 Katherine Bayless: Right.
548 01:09:12.289 ⇒ 01:09:21.829 Katherine Bayless: Yeah, cool. Well, I know we’re a little bit over time, so sorry for that, but I think this is all really exciting, and I feel like we’re in really good shape, so…
549 01:09:22.700 ⇒ 01:09:29.529 Uttam Kumaran: Yeah, so I think I’m gonna try to get, like… I’m gonna try to push what I can and think about, like, how I can create an environment for our team to test.
550 01:09:31.060 ⇒ 01:09:41.539 Uttam Kumaran: I’m still figuring that out, so give me… so give me a sec. And then, yeah, I would love to drive towards, like, the end of the month to, like, try to get something out there, for sure.
551 01:09:42.640 ⇒ 01:09:48.890 Kyle Wandel: And we gotta encourage Catherine, or you tell me… we gotta encourage Catherine to get on the stage and have a all-hands meeting so that we can…
552 01:09:48.899 ⇒ 01:09:52.519 Uttam Kumaran: Yes! What the heck? Is that even on the table? For sure.
553 01:09:52.520 ⇒ 01:09:55.250 Kyle Wandel: I want it to be. Like, I’d be pretty cool to get up there and just kind.
554 01:09:55.250 ⇒ 01:10:01.070 Uttam Kumaran: I was gonna say, like, Email is one thing, but, like, we should do… we gotta do a video.
555 01:10:01.470 ⇒ 01:10:06.290 Uttam Kumaran: Like, I think, yeah, you want to offer trainings, like, I think so too, for sure.
556 01:10:06.520 ⇒ 01:10:11.800 Uttam Kumaran: Because not only are we doing the basics, but, like, this is, I think, like, another huge piece, you know?
557 01:10:12.410 ⇒ 01:10:15.990 Kyle Wandel: I definitely think we should do, like, a lunch and learn, for sure. I think it’s a good way to kick it off.
558 01:10:16.720 ⇒ 01:10:24.770 Katherine Bayless: Yeah, and we do have budget for something like that, if we wanted to, like, you know, provide… I mean, I don’t think we could do a full meal, perhaps, but at least, you know
559 01:10:24.940 ⇒ 01:10:26.830 Katherine Bayless: Snacks. Cookies, gifts.
560 01:10:26.830 ⇒ 01:10:30.809 Uttam Kumaran: Yeah, and if you tell me, I could try to come by, if it’s in April.
561 01:10:30.970 ⇒ 01:10:37.539 Uttam Kumaran: That could be fun. I would love to come meet people and, like, do in-person training. More than happy.
562 01:10:37.910 ⇒ 01:10:39.419 Katherine Bayless: Yeah, we could totally try to coordinate.
563 01:10:39.420 ⇒ 01:10:46.050 Uttam Kumaran: I have to visit my sister soon, anyways, in DC, so… I, it’d be a good two-for-one.
564 01:10:46.050 ⇒ 01:10:47.310 Katherine Bayless: Yeah, yeah, yeah.
565 01:10:48.150 ⇒ 01:10:53.920 Katherine Bayless: Actually, I mean, we should, yeah, we should try and coordinate that, because that could be really interesting. I’m also, like.
566 01:10:54.630 ⇒ 01:11:07.380 Katherine Bayless: I don’t know, probably getting, too ahead of things, but I’m like, it’d be interesting, actually, if you came during the, like, CES Tech Week stuff, where there’s, like, a lot of activity going on generally, but then on the other hand, all of the potential users will be distracted.
567 01:11:09.010 ⇒ 01:11:18.650 Uttam Kumaran: Probably better to come before that. Yeah, I just think sometimes it’s, like, helpful for people to know that we’re trying to do stuff for them, and, like, they know me if they’re gonna be interacting with our team, and…
568 01:11:19.290 ⇒ 01:11:21.249 Uttam Kumaran: Yeah, that’d be great.
569 01:11:21.320 ⇒ 01:11:21.990 Katherine Bayless: Yeah.
570 01:11:21.990 ⇒ 01:11:27.739 Kyle Wandel: Tech Week Tech Week isn’t too bad, it’s just the GLA people. The GLA people are swamped during Tech Week, everybody else is usually okay.
571 01:11:28.120 ⇒ 01:11:33.349 Katherine Bayless: True, true, true, that’s true, yeah. And actually, I think they’ll probably be some of the latest adopters to Snowflake.
572 01:11:33.350 ⇒ 01:11:33.740 Kyle Wandel: Oh, no.
573 01:11:34.380 ⇒ 01:11:36.949 Katherine Bayless: We just don’t really have much to offer them.
574 01:11:37.280 ⇒ 01:11:41.319 Kyle Wandel: Well, as long as we don’t tell Ed anything about tariffs, we’ll be fine.
575 01:11:43.120 ⇒ 01:12:00.610 Kyle Wandel: We basically had to build out this entire tariff report for the forecast team, for our policy analyst, and obviously that data is just an absolute mess, but he’s also very excited and gung-ho, so, like, every day he would message us, how is it going? How’s it going? I’m like, dude, I don’t… there’s so much going on right now, I’m trying to keep my head above water.
576 01:12:02.170 ⇒ 01:12:07.279 Katherine Bayless: I mean, just wait until they ask for a tariff refund report or something.
577 01:12:07.280 ⇒ 01:12:11.279 Kyle Wandel: Yeah, he wants to do another one. MRD’s trying to do it right now, so…
578 01:12:12.040 ⇒ 01:12:12.760 Katherine Bayless: Yeah.
579 01:12:15.970 ⇒ 01:12:16.620 Kyle Wandel: Dear.
580 01:12:19.050 ⇒ 01:12:22.199 Uttam Kumaran: Okay, great. Alright, so then we’ll come Monday with some…
581 01:12:22.620 ⇒ 01:12:26.819 Uttam Kumaran: Prep on talking through the remaining two-week plan for this stuff, so perfect.
582 01:12:27.180 ⇒ 01:12:27.840 Katherine Bayless: Okay.
583 01:12:28.150 ⇒ 01:12:32.679 Uttam Kumaran: In a way, Sha, if you want to send the updated docs from the data modeling stuff, that’d be great.
584 01:12:33.380 ⇒ 01:12:34.090 Katherine Bayless: Oh, yeah, yeah.
585 01:12:34.480 ⇒ 01:12:41.790 Kyle Wandel: Yeah, and like I said, the only… really quickly, the only two changes I said were the ones that I commented on, which, but we’re… I think we figured those both out, so…
586 01:12:42.940 ⇒ 01:12:43.570 Uttam Kumaran: Okay.
587 01:12:44.320 ⇒ 01:12:45.340 Uttam Kumaran: Nice. Perfect.
588 01:12:47.810 ⇒ 01:12:48.620 Uttam Kumaran: Okay.
589 01:12:48.730 ⇒ 01:12:50.180 Uttam Kumaran: Thanks, everyone. Appreciate it.
590 01:12:50.750 ⇒ 01:12:52.169 Uttam Kumaran: Have a great weekend.
591 01:12:52.520 ⇒ 01:12:53.369 Uttam Kumaran: Yeah, okay.
592 01:12:53.370 ⇒ 01:12:54.040 Katherine Bayless: Thank you, bye.