Meeting Title: Brainforge x CTA: Weekly! Date: 2026-04-03 Meeting participants: Kyle Wandel, Uttam Kumaran, Chi Quinn, Ashwini Sharma, Katherine Bayless, Awaish Kumar
WEBVTT
1 00:00:10.120 ⇒ 00:00:10.900 Uttam Kumaran: Hey, Carav.
2 00:00:11.300 ⇒ 00:00:12.490 Kyle Wandel: Good morning, Dom, how you doing?
3 00:00:12.490 ⇒ 00:00:13.900 Uttam Kumaran: Hey, good, how are you?
4 00:00:13.900 ⇒ 00:00:15.089 Kyle Wandel: How was your, trip?
5 00:00:15.960 ⇒ 00:00:22.220 Uttam Kumaran: It was good! Well, I was in… I was in New York, I’m actually now in California, so it’s been a.
6 00:00:22.220 ⇒ 00:00:22.730 Kyle Wandel: Full of a bit of a.
7 00:00:22.730 ⇒ 00:00:24.820 Uttam Kumaran: Crazy.
8 00:00:24.920 ⇒ 00:00:32.420 Uttam Kumaran: travel week, I’m… my… I’m visiting family this week here, and we’re actually going, we’re going camping in Big Sur later today, so…
9 00:00:33.080 ⇒ 00:00:38.270 Kyle Wandel: That’s pretty sweet. We’re in, California, I guess. I guess I probably should know where Big Sur is, but I have no idea.
10 00:00:45.490 ⇒ 00:00:47.369 Uttam Kumaran: Did I lose you? Maybe.
11 00:00:47.590 ⇒ 00:00:50.779 Kyle Wandel: No, yeah, I think you’re… I… I taught… er, can you hear me now?
12 00:00:55.840 ⇒ 00:00:57.150 Kyle Wandel: Can you hear me now?
13 00:00:57.940 ⇒ 00:00:59.789 Kyle Wandel: I think… Okay, cool.
14 00:00:59.790 ⇒ 00:01:00.899 Chi Quinn: Hey, what’s good.
15 00:01:01.710 ⇒ 00:01:02.530 Uttam Kumaran: Hey, okay.
16 00:01:04.739 ⇒ 00:01:07.319 Kyle Wandel: Yeah, I think he was… I think he was timing out.
17 00:01:07.700 ⇒ 00:01:09.370 Chi Quinn: Oh, -oh.
18 00:01:12.970 ⇒ 00:01:13.580 Kyle Wandel: Better?
19 00:01:14.210 ⇒ 00:01:15.030 Uttam Kumaran: I don’t know if I…
20 00:01:15.480 ⇒ 00:01:17.109 Kyle Wandel: No worries. Good morning, Yoshpini.
21 00:01:17.530 ⇒ 00:01:17.950 Chi Quinn: Morning!
22 00:01:17.950 ⇒ 00:01:18.680 Ashwini Sharma: Good morning.
23 00:01:19.750 ⇒ 00:01:22.119 Kyle Wandel: So, where in California are you, Tom?
24 00:01:22.120 ⇒ 00:01:29.379 Uttam Kumaran: I’m in the East Bay right now, and I’m going… we’re going camping in Big Sur, actually, which is, like, in Northern California.
25 00:01:29.880 ⇒ 00:01:33.870 Uttam Kumaran: So take a look, it’s a big SUR, it’s like this,
26 00:01:34.240 ⇒ 00:01:36.480 Uttam Kumaran: Really, really beautiful area of California.
27 00:01:36.780 ⇒ 00:01:37.650 Chi Quinn: It is.
28 00:01:39.790 ⇒ 00:01:42.000 Kyle Wandel: Wow, oh yeah, the coast, it’s beautiful.
29 00:01:42.970 ⇒ 00:01:50.139 Uttam Kumaran: Yeah, so it’s like, we’re camping there today, but yeah, it’s just, like, right on the coastline, and should be really, really nice.
30 00:01:52.860 ⇒ 00:01:54.549 Kyle Wandel: Yeah, that looks really nice.
31 00:01:59.090 ⇒ 00:02:00.859 Chi Quinn: How long will you be camping there?
32 00:02:01.560 ⇒ 00:02:07.569 Uttam Kumaran: Just tonight. We usually do, like, one or two nights, but we’re just going tonight because we’re flying out on Sunday.
33 00:02:07.920 ⇒ 00:02:19.790 Uttam Kumaran: So we’re gonna… we’ll go there this afternoon, and then camp out, tonight, and then spend the day tomorrow, like, hiking, and eating, and hanging out.
34 00:02:21.190 ⇒ 00:02:22.159 Chi Quinn: That was nice.
35 00:02:23.550 ⇒ 00:02:25.790 Kyle Wandel: Are you religious then, you’re done?
36 00:02:25.790 ⇒ 00:02:28.660 Uttam Kumaran: Yeah, I grew up here in the Bay Area.
37 00:02:34.500 ⇒ 00:02:37.639 Kyle Wandel: Yeah, I feel like going from Texas to…
38 00:02:39.060 ⇒ 00:02:43.399 Kyle Wandel: New York all the way back to California is a lot. You’re just doing a big triangle, literally.
39 00:02:43.650 ⇒ 00:02:50.750 Uttam Kumaran: It’s a lot, yeah. The New York trip was a little bit unexpected. It was not… my intent was not to put… not to do that the same
40 00:02:51.130 ⇒ 00:02:54.520 Uttam Kumaran: this, but… You know, just lock in.
41 00:02:58.050 ⇒ 00:03:06.589 Kyle Wandel: Catherine should be joining, I haven’t heard anything of her not joining. She has some, I think, good news, so… I don’t know if she’s mentioned anything yet, but I’ll ping her.
42 00:03:06.830 ⇒ 00:03:07.450 Uttam Kumaran: Okay.
43 00:03:07.770 ⇒ 00:03:10.859 Uttam Kumaran: How was, how was your guys’ weeks, or what’s the weekend planned?
44 00:03:15.530 ⇒ 00:03:17.150 Kyle Wandel: I’m pretty good. Go ahead, Kyle.
45 00:03:18.260 ⇒ 00:03:20.300 Chi Quinn: Yeah, just… pretty much just…
46 00:03:20.420 ⇒ 00:03:33.829 Chi Quinn: I don’t know if it’s, like, work-related or personal-related, but I guess for weekend purposes, I know I’m trying to relax. I’m… well, actually, I was invited to go to a surprise party tomorrow, and I was also volunteered to apparently
47 00:03:34.460 ⇒ 00:03:39.680 Chi Quinn: I don’t know, it was… it was something. It’s something someone texted me at the last minute, I guess, to be a…
48 00:03:39.820 ⇒ 00:03:43.479 Chi Quinn: bottle girl who passed out a bottle or something, but…
49 00:03:43.940 ⇒ 00:03:44.320 Uttam Kumaran: Nice!
50 00:03:44.320 ⇒ 00:03:45.229 Chi Quinn: to work.
51 00:03:45.230 ⇒ 00:03:45.989 Uttam Kumaran: Is it a friend?
52 00:03:46.410 ⇒ 00:04:00.440 Chi Quinn: It’s a friend, yeah, so it’s supposed to be, like, a surprise party, and her and her family, they’re actually leaving the country in July, so this is kind of like a surprise birthday slash farewell type thing.
53 00:04:01.170 ⇒ 00:04:01.980 Uttam Kumaran: Nice.
54 00:04:02.670 ⇒ 00:04:03.890 Uttam Kumaran: Oh, that should be fun.
55 00:04:04.250 ⇒ 00:04:04.880 Chi Quinn: Yeah.
56 00:04:08.830 ⇒ 00:04:16.509 Kyle Wandel: I went to, the Elite Eight game last week for… because it was in DC. My wife’s a big, Duke alumni, so that was…
57 00:04:16.519 ⇒ 00:04:18.779 Uttam Kumaran: Wow, awesome game.
58 00:04:18.779 ⇒ 00:04:38.029 Kyle Wandel: It was fun, like, oh yeah, it was real fun, and like, obviously at the second… after the end of the first half, we’re like, okay, pretty, we got this, they got it. And then, I think it was, like, the first, like, 3 minutes in the second half, and they had, like, 4 fouls, and I, like, looked over the recording, and I’m like, this is… this is not good. Like, this is not gonna end well.
59 00:04:38.029 ⇒ 00:04:42.499 Kyle Wandel: And then, lo and behold, slowly but slowly, they lost, which was crazy.
60 00:04:42.500 ⇒ 00:04:47.280 Uttam Kumaran: Hey, you guys are… you’re an awesome team anyway, so, like…
61 00:04:47.280 ⇒ 00:04:47.710 Kyle Wandel: Oh, yeah.
62 00:04:47.710 ⇒ 00:05:01.769 Uttam Kumaran: be back, I’m sure, very quickly, but that’s such a clutch game to go to. That must have been so insane. I’ve never gone to see college basketball, let alone… like, I went to Bucknell, and so we were in…
63 00:05:01.900 ⇒ 00:05:06.640 Uttam Kumaran: We were in a tournament, like, a few years ago, and then also, like, maybe, like, 15 years ago, but…
64 00:05:07.110 ⇒ 00:05:10.980 Uttam Kumaran: I’ve just been to NBA games, but that environment seems crazy.
65 00:05:11.300 ⇒ 00:05:31.260 Kyle Wandel: Yeah, that was my… I mean, we… it was, like, a last-minute type thing, because we were, like, really thinking about it, and then we kind of, like, took the chance, because we bought the tickets before Duke won, and so, like, we were hoping they made it, so they made it. But then, yeah, it was just wild, to have, like, the last shot go in like that. I remember…
66 00:05:31.490 ⇒ 00:05:45.120 Kyle Wandel: Obviously, it was on the opposite end of us, we didn’t get, like, a fantastic shot, but I remember when they inbound the ball, and it happened so quickly, but I was like, oh, you gotta throw it, pass it, pass it, and then you just pass it right to the other team, like, okay, well, it’s not good.
67 00:05:47.940 ⇒ 00:05:54.089 Uttam Kumaran: I just yell at… I just… whenever I watch college, I just, like, yell at my TV screen.
68 00:05:54.240 ⇒ 00:06:05.349 Kyle Wandel: Yeah. Well, and I hate watching, like, I feel like I can only really watch tournament games now for college, because, like, in my opinion, the basketball’s, like, so bad, it’s only good during the tournament, and so it’s hard for me to watch the actual games.
69 00:06:07.240 ⇒ 00:06:08.720 Uttam Kumaran: Yeah, I agree.
70 00:06:09.190 ⇒ 00:06:10.209 Kyle Wandel: Good morning, Catherine.
71 00:06:10.210 ⇒ 00:06:15.029 Katherine Bayless: Good morning. Sorry for being a few minutes late. Joe, Joe, listen to me.
72 00:06:15.030 ⇒ 00:06:37.220 Katherine Bayless: Jay wanted to set up Snowflake, so Jay plus Snowflake became Joe. He wanted to set up, Snowflake, in… he’s actually gonna use the sandbox one, after all, to connect to the data from Concur and Ironclad, and I think maybe Profix, if possible, because he wants to, take a stab at actually trying to start reconciling some of that financial data, and so I was…
73 00:06:37.220 ⇒ 00:06:46.380 Katherine Bayless: trying to get him set up with the, like, you know, Snow CLI, kind of like we have in Claude, but, like, Kyle and I both yesterday, and me again this morning, just now, like.
74 00:06:46.650 ⇒ 00:07:01.410 Katherine Bayless: Claude seems to get oddly cranky about finding it, like, every time I start a session and I’m like, you know, connect to Snowflake, it’s like, oh, you don’t have it installed, let me try again. I think I finally figured out where it was getting stuck, that there was, like.
75 00:07:01.540 ⇒ 00:07:21.270 Katherine Bayless: the path reference was in my ZSHRC, but not my Z profile. So, I don’t know. Anyway, that was all. Just… I was like, I refuse to switch calls until I actually get this thing. So, he’s off and running. He’s also off today, but, can’t resist playing with data.
76 00:07:21.590 ⇒ 00:07:23.490 Uttam Kumaran: That’s great.
77 00:07:24.000 ⇒ 00:07:29.139 Katherine Bayless: So yeah, so we’ll see what he comes back with, but… Happy Friday, everybody!
78 00:07:29.610 ⇒ 00:07:30.490 Uttam Kumaran: Friday.
79 00:07:30.670 ⇒ 00:07:31.610 Chi Quinn: Yep, ready.
80 00:07:32.160 ⇒ 00:07:35.290 Katherine Bayless: I don’t know about y’all, but I am, I am…
81 00:07:36.020 ⇒ 00:07:39.109 Katherine Bayless: My brain is, like, totally fried.
82 00:07:39.770 ⇒ 00:07:41.359 Kyle Wandel: That’s how it was last night, yeah.
83 00:07:41.360 ⇒ 00:07:43.030 Katherine Bayless: Yeah? Yeah.
84 00:07:43.490 ⇒ 00:07:54.490 Uttam Kumaran: Yeah, it’s been a busy week, and then I’ve been traveling a little bit, so it’s been a little bit all over. But really good. But I also agree, it’s been really positive. I feel like we made a lot of progress, and I think,
85 00:07:54.690 ⇒ 00:08:00.359 Uttam Kumaran: I feel like Awash has, should we have some stuff to share on modeling side, so she’ll… I feel pretty good.
86 00:08:00.890 ⇒ 00:08:14.459 Katherine Bayless: Cool. I’ve got some, admin-y updates, and then some data updates, and then maybe we can talk about the Power BI stuff a little bit, too. But… and then, I don’t know, Kyle or Kai, if you’ve got stuff we want to look at as well.
87 00:08:15.620 ⇒ 00:08:16.239 Kyle Wandel: Yeah, I guess.
88 00:08:16.240 ⇒ 00:08:17.700 Katherine Bayless: But the ExpoCats style.
89 00:08:17.890 ⇒ 00:08:18.610 Kyle Wandel: What?
90 00:08:18.610 ⇒ 00:08:20.339 Katherine Bayless: We can talk a little bit about the ExpoCAD stuff, Kyle.
91 00:08:20.340 ⇒ 00:08:35.520 Kyle Wandel: Yeah, I mean, I mean, I could… the stuff that I’ve done, like, was in that summary that I kind of sent last night, most of it. I don’t think I included the ExpoCAD kind of Python, but, I mean, do we want… do you want to do, like, a round robin, or do you just want to kind of, like, go over what we’ve done?
92 00:08:36.830 ⇒ 00:08:38.820 Katherine Bayless: Let’s go round robin. Wanna go first?
93 00:08:39.329 ⇒ 00:08:53.289 Kyle Wandel: Sure, as always. So, I mean, like, the biggest thing I kind of want to do, and again, I cleaned up a lot of the PRs, I started going through, like, maybe, like, 6 or 7 at the time, so I started cleaning up some of the PRs, merged the star schema, so I think that should be good to go.
94 00:08:53.349 ⇒ 00:09:00.069 Kyle Wandel: I did do some cleanup of the badge cans a little bit, mainly trying to… I think we want to try and…
95 00:09:00.169 ⇒ 00:09:16.199 Kyle Wandel: get rid of the member engagement kind of schemas, because we used those initially for the member engagement report, but… but now that we have a more working dbt model and a more star schema built out, I think we can probably start to truncate those, or whatever, make them better.
96 00:09:16.569 ⇒ 00:09:33.119 Kyle Wandel: And so I was kind of doing that here and there, and kind of going through some of the member engagement tables, deleting them if need be, creating new dbt pipelines if need be for them. And then the one big thing that I definitely probably need to review, and I’m not 100% sure if I was doing it.
97 00:09:33.119 ⇒ 00:09:37.249 Kyle Wandel: Correctly, was, like, the role permissions for…
98 00:09:37.249 ⇒ 00:09:38.989 Kyle Wandel: A lot of the roles.
99 00:09:39.019 ⇒ 00:09:46.609 Kyle Wandel: I don’t know if it was because we created them earlier or what, but a lot of us didn’t have… I didn’t have permissions especially on those member engagement schemas.
100 00:09:46.739 ⇒ 00:09:52.579 Kyle Wandel: And so I kind of tweaked some stuff. Again, I don’t know if that tweaking was great or not, but I did it, so…
101 00:09:52.919 ⇒ 00:09:58.969 Kyle Wandel: I’m happy to chat, or have it QA’d, or changed back, or whatever, but, that’s what I did.
102 00:09:59.370 ⇒ 00:10:01.560 Uttam Kumaran: Okay. I can take a look at that today.
103 00:10:03.560 ⇒ 00:10:15.700 Awaish Kumar: Yeah, I looked at that PR, but, like, you’re granting the CICD role access to, like, product staging database and things like that. Yep.
104 00:10:16.940 ⇒ 00:10:29.449 Awaish Kumar: But, yeah, I… I don’t know why it was failing for you, because if it didn’t have access, it could… it failed for all the PRs, because product staging is basically
105 00:10:30.080 ⇒ 00:10:34.910 Awaish Kumar: kind of has all the SDG models, so it should have been failing for all.
106 00:10:35.110 ⇒ 00:10:39.499 Awaish Kumar: Yeah, but yeah, I can look deeper into that.
107 00:10:39.500 ⇒ 00:10:44.620 Kyle Wandel: Yeah, and I think it was failing for a while, I think.
108 00:10:44.750 ⇒ 00:10:59.499 Kyle Wandel: But then when I made the changes, it was able to run. I mean, I think the biggest thing was I wasn’t able to modify existing schemas. I think that was the biggest issue. Not necessarily creating new schemas, but it was more so the ability to modify current ones.
109 00:10:59.720 ⇒ 00:11:05.470 Kyle Wandel: I think that was the biggest issue. I think there were… like, I don’t think it was prod…
110 00:11:06.210 ⇒ 00:11:09.999 Kyle Wandel: I see ProdWright over there, but I think it would just have to do…
111 00:11:12.550 ⇒ 00:11:21.580 Kyle Wandel: I’ll have to look, but now I see that. But I don’t know why it was not letting me do it, but I was trying to, like, bash through it, and then eventually when I made those changes, it worked, so…
112 00:11:22.050 ⇒ 00:11:22.700 Awaish Kumar: Okay.
113 00:11:24.150 ⇒ 00:11:29.820 Awaish Kumar: And yeah, apart from that, I’ve also been like.
114 00:11:31.050 ⇒ 00:11:35.260 Awaish Kumar: giving the, grants to, for example, Snowflake.
115 00:11:35.630 ⇒ 00:11:38.480 Awaish Kumar: Request viewer, so that,
116 00:11:38.710 ⇒ 00:11:46.530 Awaish Kumar: to CCD role so that it doesn’t fail the PRs. So, all these grants, I think now it should be in good place.
117 00:11:48.930 ⇒ 00:11:58.129 Awaish Kumar: Apart from that, I’ve also been trying to clean up the PRs and try to merge all the work we have done so far.
118 00:11:58.320 ⇒ 00:12:01.639 Awaish Kumar: So it gets to production, and we start using it.
119 00:12:03.260 ⇒ 00:12:09.350 Awaish Kumar: After that, I’ve been working with building snow pipes for the SFMC.
120 00:12:09.480 ⇒ 00:12:17.820 Awaish Kumar: data, yeah, and trying to figure out, automating the ways to create those snow piles, because
121 00:12:18.090 ⇒ 00:12:24.810 Awaish Kumar: The manual way is… takes some time to basically go through each different path.
122 00:12:25.250 ⇒ 00:12:33.209 Awaish Kumar: And… And after that, I’ve been modeling it, so… So, like, right now.
123 00:12:33.360 ⇒ 00:12:39.130 Awaish Kumar: I have modeled it based on whatever I… I could read about SOMC.
124 00:12:39.280 ⇒ 00:12:40.000 Awaish Kumar: Hmm.
125 00:12:40.100 ⇒ 00:12:43.670 Awaish Kumar: But, like, we don’t have exact requirements.
126 00:12:43.800 ⇒ 00:12:56.840 Awaish Kumar: Like, what… at the end, what are the end goal? What are the questions that we want to have answers for? So we have some standard DIM and fact tables, which can give you, like, subscribers, which connects to…
127 00:12:57.100 ⇒ 00:13:04.259 Awaish Kumar: Sends, which are what… what email templates are good sends, for whom, like, things like that, but…
128 00:13:04.490 ⇒ 00:13:08.660 Awaish Kumar: Then we might work on creating somebody tables if needed.
129 00:13:08.870 ⇒ 00:13:12.529 Awaish Kumar: If we have, like, the exact business questions.
130 00:13:13.320 ⇒ 00:13:19.430 Katherine Bayless: Yeah, actually, that’s, that’s a great kind of segue, Oish, because I think,
131 00:13:19.430 ⇒ 00:13:39.739 Katherine Bayless: Yeah, I mean, Salesforce Marketing Cloud’s a good example of, like, there’s the obvious use case of, like, we want to see engagement with the emails, right? Like, is this, you know, user, or not user, this individual that we’ve got, you know, as a company’s primary rep, ignoring us all the time? Is it bouncing all the time? Are they reading them? Like, you know, so the low-hanging fruit, obvious stuff.
132 00:13:39.740 ⇒ 00:14:04.480 Katherine Bayless: But I think there are probably other questions and things we could answer out of that data that would be helpful guidance. And then I’m also thinking ahead to what I was going to talk about with some of the other data I’ve been landing, where, yeah, you would probably be like, I don’t know what you want to do with this. So I can put in, like, Asana tickets with the data that’s landed, and then kind of, like, these are the questions we’re looking at answering with it, like, right away, and then, you know, whatever else interesting stuff turns out
133 00:14:04.480 ⇒ 00:14:09.419 Katherine Bayless: up fine. But yeah, I could totally give you some of that kind of as jumping off points.
134 00:14:11.410 ⇒ 00:14:18.620 Awaish Kumar: Yeah, basically, it’s… it’s more like, like, we have all this data that you just talked about, like, we have all the events
135 00:14:18.770 ⇒ 00:14:32.439 Awaish Kumar: For each individual, we have subscriber keys to figure out which user it belongs to, and then we can connect to the exact sends, and it can give us the opens, clicks, and all that information.
136 00:14:32.480 ⇒ 00:14:40.139 Awaish Kumar: Yeah. So, all… they are… all these tables are connected via foreign keys, so…
137 00:14:40.210 ⇒ 00:14:47.939 Awaish Kumar: Coco should be able to answer it if… yeah, but we could kind of stitch together some tables if needed.
138 00:14:49.330 ⇒ 00:14:55.999 Katherine Bayless: Yeah, okay, okay, cool. I will go through, and I’ll put together the list of the business questions we’re looking at.
139 00:14:56.280 ⇒ 00:15:09.889 Katherine Bayless: Yeah. Yeah. I did notice, when I finally… the Fivetran connector finished syncing for Marketing Cloud, I think it was, like, 30 or so million rows, of stuff that came back, so, yeah, lots of… lots of stuff to play with in there.
140 00:15:09.890 ⇒ 00:15:11.080 Uttam Kumaran: Yeah. Great.
141 00:15:11.270 ⇒ 00:15:12.720 Katherine Bayless: Yeah, yeah.
142 00:15:14.550 ⇒ 00:15:26.470 Katherine Bayless: Actually, I mean, maybe that’s a decent segue, just while we’re on the topic of landing more data. Let me go back over here. Okay, so I… obviously, I set up Marketing Cloud.
143 00:15:26.470 ⇒ 00:15:37.900 Katherine Bayless: I also set up, Qualtrics, which is the survey tool that Market Research is using, and so that data is in that, raw Fivetran, then Qualtrics.
144 00:15:37.900 ⇒ 00:16:02.119 Katherine Bayless: directory as well. I only had access to one of the surveys, so, maybe if we can convince market research to let a little bit more data flow through. But in the meantime, at the very least, like, the structure of the Qualtrics data is in there, so we can start modeling it, and then if another survey came through, we would just… it’d be a matter of, like, saying, now, give me, you know, that survey ID, kind of instead. So it’s a much smaller body of data overall.
145 00:16:02.370 ⇒ 00:16:05.460 Katherine Bayless: I also am…
146 00:16:06.670 ⇒ 00:16:24.179 Katherine Bayless: I think I gave up on the Fivetran connector for Cvent, because we have the Cvent tier that’s, like, you get, like, a thousand calls a day, and, like, getting a max of, like, two per minute or something, and the Fivetran connector just could not get out of its own way, trying to pull the data down, and so it was just, like, getting, like, rate-limited and blocked constantly.
147 00:16:24.180 ⇒ 00:16:28.509 Katherine Bayless: So, I think I managed to get it all written with Cloud Code.
148 00:16:30.600 ⇒ 00:16:49.120 Katherine Bayless: I have to admit, every now and then, the robot really drives me fucking crazy, because I got it all working last night, and it’s like, the one piece of the data that we had to load incrementally to play nice with the API was the attendees, and so I was like, well, let’s just batch them. I’ll pull one year of attendees, and then the next year, and then the next year, and then we’ll have the initial load done, it’ll be fine.
149 00:16:49.340 ⇒ 00:16:57.690 Katherine Bayless: I didn’t pay attention to the fact that it was not naming the files in a way that would prevent overwrites, so basically I pulled all of the Cvent data, and then I overwrote all of it.
150 00:16:57.830 ⇒ 00:16:59.140 Katherine Bayless: So, we’ll pull it again today.
151 00:16:59.140 ⇒ 00:17:01.300 Uttam Kumaran: Ugh.
152 00:17:04.040 ⇒ 00:17:05.490 Uttam Kumaran: Nice.
153 00:17:05.700 ⇒ 00:17:19.489 Katherine Bayless: Yeah. So yeah, so this event data by the end of today, will be in there. That’s gonna be one that’s interesting, I think, to deal with. Like, it’s a very sprawling kind of data model.
154 00:17:19.650 ⇒ 00:17:28.510 Katherine Bayless: I’ve worked with Cvent in the past, but not with as, like, robust of a use case. Like, I think we actually do use a lot of the features in there, so…
155 00:17:28.520 ⇒ 00:17:41.100 Katherine Bayless: Yeah, it’ll be interesting to kind of play with. But similarly, I can define kind of, like, the business questions we’re looking for, like, right away versus, you know, later if we find interesting things we can play with. Kyle, did you have… you wanted to add something?
156 00:17:41.100 ⇒ 00:17:55.309 Kyle Wandel: I was just gonna say, I mean, it’s a pretty important data set, too. I mean, it’s something I have never played with, but I also know that it’s people, it’s pretty coveted by most departments, especially for, like, briefers, executive reports, membership, so it’s a pretty nice dataset.
157 00:17:55.730 ⇒ 00:18:12.360 Katherine Bayless: Yeah, and actually, as it happens, Christine Michael, who’s the, like, you know, person who runs the team that uses this Cvent instance, she was super excited to hear that it would be in Snowflake. So, in that sense, we don’t have any sort of, like, you know, data fear to worry about, as we start to make it more available. Go ahead.
158 00:18:12.360 ⇒ 00:18:24.740 Kyle Wandel: So, because that API is going to be historical, do we need to go through the DBeaver or old marketing stuff to see it and pull other events? So, like, we have it in S3 already, I think it’s other events is the bucket.
159 00:18:26.900 ⇒ 00:18:31.549 Kyle Wandel: So I don’t know if we… we probably will obviously keep it, obviously, I’m not saying that, but I’m saying, do we need to…
160 00:18:31.800 ⇒ 00:18:35.619 Kyle Wandel: Use that data, or are we basically just gonna rebuild it from scratch?
161 00:18:36.670 ⇒ 00:18:37.529 Kyle Wandel: Question I would have.
162 00:18:37.770 ⇒ 00:18:40.560 Katherine Bayless: Yeah, that’s a great question. I think…
163 00:18:41.360 ⇒ 00:18:48.870 Katherine Bayless: I think probably, I guess, the first step is making sure that everything that I can pull through Cvent, because it will pull everything in the account.
164 00:18:48.870 ⇒ 00:19:10.450 Katherine Bayless: But if people have, like, deleted stuff in the account across the years, we might have more data in some of the old stuff, so it’s probably worth doing a one-pass, at least, like, audit to make sure that, like, there’s not anything more in the old stuff that we would potentially wind up losing. But then, yeah, once we validate that all the data is in both, I think we probably should move…
165 00:19:10.450 ⇒ 00:19:22.070 Katherine Bayless: the old data sources, like, just… I mean, they’re already in that archive bucket, necessarily, because we kind of had copied them into this one, but yeah, I think we should get them out of the data lake so that they don’t… don’t confuse people, honestly.
166 00:19:22.070 ⇒ 00:19:38.649 Katherine Bayless: The same as, like, I need to go delete the data that Fivetran did pull down for Cvent, because then, right, that’s another layer of confusion. So, I think, yeah, I think the cleanup once we start consuming these data sources programmatically, is a good use of our time, just to make sure that we’re not
167 00:19:39.110 ⇒ 00:19:42.659 Katherine Bayless: Keeping two similar-ish copies of data running around.
168 00:19:44.340 ⇒ 00:19:53.120 Kyle Wandel: Yeah, and I think we were pretty close to that, and it goes back to me trying to clean up the member engagement stuff, too, so I think, whenever we want to start doing that, we certainly… I think that was going to be one of my next big
169 00:19:53.720 ⇒ 00:20:01.350 Kyle Wandel: goals, but attempts, is once we get the star schemas all approved, the ExpoCAD stuff all approved, I was gonna go in and clean a little bit.
170 00:20:01.740 ⇒ 00:20:04.389 Katherine Bayless: Yeah, I like that. I like that.
171 00:20:06.720 ⇒ 00:20:09.849 Awaish Kumar: What is this, like, CVAN data, and…
172 00:20:10.840 ⇒ 00:20:28.650 Katherine Bayless: It’s, so it’s an event management platform, and it does everything from, like, you can make, like, a marketing website for your event, to, like, a seating chart for the tables, and everything in between. We use it for some of our, like, smaller events, like, we do the Digital Patriots Dinner, CES on the Hill.
173 00:20:29.000 ⇒ 00:20:40.300 Katherine Bayless: other things, CEO Summit, and so for us, we’re running registration through Cvent for anything that’s not CES, basically, or just a webinar. Webinars are in Zoom, which also working on.
174 00:20:41.880 ⇒ 00:20:44.789 Awaish Kumar: So it’s event management, like, platform.
175 00:20:45.940 ⇒ 00:20:46.730 Awaish Kumar: Basically.
176 00:20:46.730 ⇒ 00:20:51.039 Kyle Wandel: Basically, all the data around all the events we have that is not CES.
177 00:20:51.330 ⇒ 00:20:51.740 Katherine Bayless: Hmm.
178 00:20:52.150 ⇒ 00:20:53.049 Awaish Kumar: Okay, okay.
179 00:20:53.930 ⇒ 00:21:01.800 Katherine Bayless: I think the one interesting, like, modeling wrinkle that we’ll run into, not immediately, but, like, eventually, is…
180 00:21:01.820 ⇒ 00:21:16.479 Katherine Bayless: a lot of our Cvent events are gated in some way, like, not just anybody… I mean, in theory, the same with CES, but more so for these events, right? Like, you have to be kind of invited, and then you may choose to attend or not, right?
181 00:21:17.130 ⇒ 00:21:29.260 Katherine Bayless: I want to be able to capture that conversion rate, but I don’t know how often we’ve put the full invitee list in Cvent so that they can, you know, choose to register or not, versus
182 00:21:29.260 ⇒ 00:21:43.719 Katherine Bayless: sent an email to the invitee list in Marketing Cloud, and just kind of controlled the access to it that way. And so it might be interesting when we get there to figure out, like, okay, we can see now who was registered and attended for these events.
183 00:21:43.720 ⇒ 00:21:55.849 Katherine Bayless: But if we want to look at the invitation side of the equation, we’ll maybe need to stitch the data between CVEN and Marketing Cloud in order to figure out that piece, but that’s not, like, a day one need, fortunately.
184 00:21:56.730 ⇒ 00:22:02.250 Awaish Kumar: By the Marketing Cloud, it does have data on, like, what… which people received that email.
185 00:22:03.430 ⇒ 00:22:08.750 Awaish Kumar: mail has its own, like, title. It’s, like, really easy to stitch together.
186 00:22:09.190 ⇒ 00:22:17.370 Katherine Bayless: Yeah, yeah, exactly. So I think if we can figure out which Cvent events had the full invitee list versus not, then maybe it won’t be too difficult to…
187 00:22:17.490 ⇒ 00:22:20.120 Katherine Bayless: pull in the Marketing Cloud ones for the ones that didn’t.
188 00:22:20.980 ⇒ 00:22:21.800 Katherine Bayless: Yeah.
189 00:22:22.380 ⇒ 00:22:29.389 Katherine Bayless: So then the other data source I started bringing down through Fivetran, or trying anyway, was Shopify.
190 00:22:31.550 ⇒ 00:22:44.060 Katherine Bayless: I’m really, like, the Shopify thing’s starting to get on my nerves, too. I don’t think the Fivetran connector… unfortunately, I initially had scoped the app on the Shopify side too tightly.
191 00:22:44.210 ⇒ 00:23:03.820 Katherine Bayless: I feel like Fivetran can’t get past its memory of the tighter scope, and, like, it just keeps hanging. Like, it’ll run for hours and pull down no data, but I can connect to the app and pull data, so it’s like, it’s… I think Fivetran’s just got something cached in there, and so I, similarly, with that one, was like, okay, well, maybe I can work around it in Cloud Code,
192 00:23:03.940 ⇒ 00:23:13.259 Katherine Bayless: But the Shopify data is another one that I have learned is being kind of fragmented across multiple platforms, so…
193 00:23:13.260 ⇒ 00:23:24.460 Katherine Bayless: we sell some of the products in Shopify, and then we also sell them in remembers. We also used to only sell them in remembers, and so there’s, like.
194 00:23:24.580 ⇒ 00:23:37.030 Katherine Bayless: almost 3 versions across two systems of this data, and so once I get a full copy of it down, it might be worth doing, like, a little spike, to kind of talk through, like, okay, what do we really need to pull through here?
195 00:23:37.100 ⇒ 00:23:48.699 Katherine Bayless: And again, I’ll put in the Asana tickets, but with this one, there is an active question coming from the Tech and Standards team around exactly what sort of haircut we took by starting to charge for the standards.
196 00:23:48.940 ⇒ 00:23:52.569 Katherine Bayless: But right now, it’s kind of impossible to answer that question, because the.
197 00:23:52.570 ⇒ 00:23:53.419 Uttam Kumaran: It is neat.
198 00:23:53.420 ⇒ 00:23:55.369 Katherine Bayless: In a few places, so…
199 00:23:55.750 ⇒ 00:23:59.109 Kyle Wandel: Who is using, Paxium still?
200 00:23:59.390 ⇒ 00:24:09.139 Kyle Wandel: when I looked earlier, I think it was… I think we haven’t had any purchases. I mean, that might be different than downloads, but we haven’t had any purchases of reports since, like, 2020, and remembers, at least.
201 00:24:09.470 ⇒ 00:24:20.890 Katherine Bayless: Yeah, apparently we started offering them in remembers again in the fall? As far as I could tell so far, there have been, I think, 8 purchase… well, purchases.
202 00:24:20.890 ⇒ 00:24:21.500 Kyle Wandel: Okay.
203 00:24:21.500 ⇒ 00:24:26.109 Katherine Bayless: It’s set up in remembers to be free for members, paid for non-members.
204 00:24:27.150 ⇒ 00:24:32.540 Katherine Bayless: I think there were two non-member transactions, but I’m like, how on earth Did that
205 00:24:32.900 ⇒ 00:24:39.980 Katherine Bayless: non-members end up buying it through the member version of the store? So those might be test data. I haven’t actually looked at them.
206 00:24:40.200 ⇒ 00:24:48.479 Kyle Wandel: I think it just speaks to broader issues of, like, our website, and our store, and how they communicate, and all that fun stuff, but… okay, interesting.
207 00:24:48.960 ⇒ 00:25:10.320 Katherine Bayless: Yeah, I mean, honestly, Kyle, to your point, I have a feeling once we get this Shopify data in hand and can start talking to it, I think the pitchforks are gonna come for that platform fast, and I’m not mad about that, right? Like, I just don’t think it works for what we’re trying to do here, and I don’t think it would be that hard to just, like, switch to using Stripe or something else, or even just
208 00:25:10.320 ⇒ 00:25:12.340 Katherine Bayless: Use remembers again, honestly.
209 00:25:12.600 ⇒ 00:25:22.239 Kyle Wandel: I mean, it’s… it’s similar to, in my opinion, it’s similar to the Badgers, the conference scans type of issues, and Kinsey put the kibosh on that real quickly, so…
210 00:25:22.550 ⇒ 00:25:26.630 Katherine Bayless: Yeah, exactly. Yeah. So I have a feeling Shopify is,
211 00:25:27.120 ⇒ 00:25:30.560 Katherine Bayless: Gonna be in the spotlight. Not in a bad way.
212 00:25:30.940 ⇒ 00:25:31.550 Kyle Wandel: Yeah.
213 00:25:32.850 ⇒ 00:25:36.729 Katherine Bayless: Yeah. Okay, and then the last one that I was able.
214 00:25:36.730 ⇒ 00:25:46.199 Uttam Kumaran: Yeah, I guess on that piece, I’m surprised, because previously, I know sometimes setting up the Shopify credentials and then redoing it, like, I’ve never seen it be sticky, so…
215 00:25:46.680 ⇒ 00:25:53.359 Uttam Kumaran: maybe worth writing a ticket to Fivetran support, maybe it’s a cash thing or something, because usually it resolves.
216 00:25:54.380 ⇒ 00:25:56.430 Katherine Bayless: Yeah, that’s true.
217 00:25:57.090 ⇒ 00:25:57.860 Katherine Bayless: I need to…
218 00:25:57.860 ⇒ 00:26:02.419 Ashwini Sharma: Best thing to do… sorry, best thing to do is delete that connection, create a new app.
219 00:26:02.800 ⇒ 00:26:03.490 Katherine Bayless: I did, yeah.
220 00:26:03.490 ⇒ 00:26:05.949 Ashwini Sharma: And then clear the connection again. That should work.
221 00:26:06.240 ⇒ 00:26:08.839 Katherine Bayless: Yeah, like, I tried tearing down, like, both sides.
222 00:26:09.180 ⇒ 00:26:09.540 Uttam Kumaran: Okay.
223 00:26:09.960 ⇒ 00:26:10.520 Katherine Bayless: seemed.
224 00:26:10.520 ⇒ 00:26:11.640 Ashwini Sharma: Still didn’t work?
225 00:26:12.690 ⇒ 00:26:16.579 Uttam Kumaran: Yeah, Ashwini actually worked for 5 train, so sorry, you shouldn’t, you shouldn’t give…
226 00:26:16.580 ⇒ 00:26:18.520 Katherine Bayless: That’s right, I forgot about that, that’s fair.
227 00:26:18.520 ⇒ 00:26:19.539 Ashwini Sharma: created that connector.
228 00:26:19.600 ⇒ 00:26:22.650 Uttam Kumaran: Oh, yeah, great, great, great.
229 00:26:23.220 ⇒ 00:26:31.380 Katherine Bayless: Oh, that’s funny. Yeah, maybe we should look at it together, Ashriti, if I can’t get it to play nice today, yeah.
230 00:26:31.740 ⇒ 00:26:44.610 Ashwini Sharma: Sure, yeah. But yeah, I mean, if the volume of data is high in Shopify, you should also add that scope to create webhooks for Shopify. Scope to read webhooks. Create webhooks as well, yeah.
231 00:26:44.940 ⇒ 00:26:48.020 Katherine Bayless: Okay, I’ll make that note. But I do think I… I don’t.
232 00:26:48.020 ⇒ 00:26:51.910 Ashwini Sharma: So that helps when the volume of transactions and orders are high.
233 00:26:52.640 ⇒ 00:26:53.250 Awaish Kumar: Yeah.
234 00:26:53.470 ⇒ 00:26:57.470 Awaish Kumar: But I don’t think so we have, like… A higher volume of…
235 00:26:57.810 ⇒ 00:27:02.230 Uttam Kumaran: Yeah, I don’t think it’s gonna be, like, millions of orders a month, you know?
236 00:27:02.640 ⇒ 00:27:03.740 Katherine Bayless: Oh god, no.
237 00:27:03.740 ⇒ 00:27:04.090 Uttam Kumaran: Okay.
238 00:27:04.090 ⇒ 00:27:07.070 Katherine Bayless: I think it’s, like, a few hundred orders ever.
239 00:27:07.570 ⇒ 00:27:09.590 Ashwini Sharma: Oh, you don’t need that.
240 00:27:09.780 ⇒ 00:27:25.279 Katherine Bayless: Oh, okay, okay, okay, okay. I thought you were saying it might help with the Fivetran thing go faster, but yeah, yeah, it’s not a ton of data. Like, I think Kyle and I might have realized, I don’t know that we always break even on what it costs to have Shopify versus what we make selling things through it.
241 00:27:26.870 ⇒ 00:27:35.129 Kyle Wandel: I didn’t even… I didn’t even know the cost of Shopify was what it was, and then Kevin told me, I was like, I think, yeah, we’re definitely not covering that, so…
242 00:27:35.300 ⇒ 00:27:37.019 Katherine Bayless: Yeah, so yeah.
243 00:27:37.150 ⇒ 00:27:41.999 Katherine Bayless: I think the platform is probably lovely, but not necessarily the right one for us.
244 00:27:44.260 ⇒ 00:28:01.640 Katherine Bayless: so anyway, so the other, the other data source that I was successful with, was, Formstack. So I actually did two things with Formstack. So one is we have a few forms in there that we’re using to essentially manage, like, newsletter interest kind of thing.
245 00:28:01.640 ⇒ 00:28:18.490 Katherine Bayless: So those, I have connected from Formstack to, sort of, like, S3, Fivetran, and then our email validation platform, and Marketing Cloud using the reverse ETL, the activations thing. And so those are just running, I think, on, like, a 15-minute cadence.
246 00:28:18.760 ⇒ 00:28:38.909 Katherine Bayless: Then, separately, I also, put the Formstack connector in to just dump everything out of Formstack into our data lake. We have a lot of stuff in there. I assume very little of it is ultimately valuable or interesting, so I wouldn’t say it’s, like, top priority, but, you know, if somebody’s got
247 00:28:38.910 ⇒ 00:28:43.570 Katherine Bayless: an hour to kill and curiosity to spare than seeing what is in Formstack?
248 00:28:43.720 ⇒ 00:29:02.960 Katherine Bayless: is now possible. But yeah, so it’s got the one directory that’s under raw Fivetran formstack is just the dump of everything out of the platform, and then separately, I have raw form stack that is where the activation stuff is going through. Probably needs slightly better information architecture eventually, but that was just where I kind of started.
249 00:29:03.510 ⇒ 00:29:07.170 Awaish Kumar: Okay, I’ve seen a PR with farmstack data, like.
250 00:29:08.480 ⇒ 00:29:11.689 Awaish Kumar: Is that something that is pulling data from Formstack?
251 00:29:11.760 ⇒ 00:29:24.879 Katherine Bayless: Yeah, so that one, actually, I’m glad you reminded me. I did put the PR in. So that, PR is for the pipelines for the reverse ETL activation. So it’s the, Lambda functions that grab the webhook, process it.
252 00:29:24.880 ⇒ 00:29:34.079 Katherine Bayless: We’ve got a little bit of, like, a DynamoDB sort of intermediary to tell us whether or not we’ve recently validated the email address, and if we haven’t, it goes through NeverBounce.
253 00:29:34.140 ⇒ 00:29:46.910 Katherine Bayless: if we have, it just pulls the data from Dynamo, and then stages it in another S3 bucket, and then the Fivetran thing picks it up on the other side. So that PR just has all of the code for, like, the Lambda functions and processing.
254 00:29:47.150 ⇒ 00:29:51.699 Katherine Bayless: I would love if somebody wants to take a look at it and let me know if it looks like a toddler wrote it.
255 00:29:55.100 ⇒ 00:30:12.100 Katherine Bayless: So yeah, so those are all the data sources that I’ve managed to land so far. I did go through and bookmark a bunch of the connectors that we have, subscriptions to the platforms for. I think there were, like, 40 or something, but I don’t necessarily have the admin access to do the connecting yet.
256 00:30:12.100 ⇒ 00:30:20.130 Katherine Bayless: For all of them, so I’m working with getting Zoom next, because that’s another place that membership needs the data from for those webinars and stuff.
257 00:30:20.180 ⇒ 00:30:31.349 Katherine Bayless: And then, Kinsey was interested in the social media, analytics, so, like, our LinkedIn ads, Instagram, blah blah blah. So I might…
258 00:30:31.350 ⇒ 00:30:41.839 Katherine Bayless: I don’t know if I’ll get to it maybe next week, but maybe a week after or so, I’ll reach out to the marketing team and say, like, I just need access to those accounts so that I can connect the APIs and start bringing in the data.
259 00:30:41.840 ⇒ 00:30:47.810 Katherine Bayless: And then we’ll, we’ll see, see how that conversation progresses. Yes.
260 00:30:47.810 ⇒ 00:30:52.790 Awaish Kumar: Okay, how you are looking to ingest that data, like, directly hitting the APIs?
261 00:30:53.480 ⇒ 00:30:59.499 Katherine Bayless: I would use the Fivetran connectors for them, so I think that’s kind of what it does under the hood, I guess?
262 00:31:01.860 ⇒ 00:31:02.470 Awaish Kumar: Okay.
263 00:31:03.170 ⇒ 00:31:03.730 Katherine Bayless: Yeah.
264 00:31:04.070 ⇒ 00:31:19.290 Katherine Bayless: And then just dumping everything into S3. I didn’t actually set up Fivetran to connect to Snowflake directly, because, I mean, it seemed kind of unnecessary if all the data needs to get into S3 kind of anyway, and then be, like, piped in with dbt into Snowflake, so that was the route I took.
265 00:31:21.310 ⇒ 00:31:21.890 Awaish Kumar: Yeah.
266 00:31:21.890 ⇒ 00:31:24.280 Kyle Wandel: Better, because we want the… we want the data storage anyways.
267 00:31:24.910 ⇒ 00:31:25.260 Awaish Kumar: Yeah.
268 00:31:25.260 ⇒ 00:31:26.890 Katherine Bayless: Yeah. Yeah.
269 00:31:28.690 ⇒ 00:31:44.070 Katherine Bayless: So yeah, so that’s the latest on fun new data landing in there. And then actually, I mean, Kyle, if you want, I’ll bounce it back to you to talk a little bit about, like, the ExpoCAD pipeline, like, in terms of, like, the application that we are working on lifting and shifting.
270 00:31:44.070 ⇒ 00:31:53.349 Kyle Wandel: Yeah, sure, I mean, very quickly, but I mean, you did a lot of the legwork in terms of, like, getting the plan ready to go, and kind of, working it over, but essentially just lifting and shifting
271 00:31:53.500 ⇒ 00:31:57.550 Kyle Wandel: a model that Jay made, to basically…
272 00:31:57.580 ⇒ 00:32:13.429 Kyle Wandel: create an exhibitor report, and we’re moving it from our other C… or whatever, CTA AWS Warehouse account to our new CTA AWS Warehouse account. And so basically just kind of going through that whole process. So yesterday, I did the initial deployment, basically setting up the infrastructure.
273 00:32:13.430 ⇒ 00:32:24.160 Kyle Wandel: for, all the functions that are needed to create it, and call it, and the secrets that need to basically store, the API call and, configurations. So I…
274 00:32:24.160 ⇒ 00:32:28.909 Kyle Wandel: just working my way through it, probably a little bit next week, just going step by step. I think there are…
275 00:32:28.910 ⇒ 00:32:35.300 Kyle Wandel: 7 total deploys, I think you have, Catherine. So I’m just working my way through that.
276 00:32:35.500 ⇒ 00:32:47.970 Kyle Wandel: But yeah, I mean, it will… hopefully we’ll set up to that… the ExpoCAD schema that you’re… that Ashwini set up, that way we can… it should be the… it should be the exact same data, or data, is my assumption. So…
277 00:32:47.970 ⇒ 00:32:50.390 Katherine Bayless: My assumption as well.
278 00:32:50.670 ⇒ 00:32:59.610 Kyle Wandel: And even me, and if it’s not, I’ll probably force it to be that way, so that won’t change any of the dbt models. So,
279 00:33:00.940 ⇒ 00:33:01.690 Kyle Wandel: That’s good.
280 00:33:02.060 ⇒ 00:33:20.019 Katherine Bayless: Yeah, actually, I’ll show, so what Kyle’s talking about with the deploy thing, so this is… and I’d be curious, honestly, Brainforge folks, to get your thoughts on this as an approach, because I started doing this with my little, like, hobby projects. Let me share my screen. And it seemed to work nicely as a sort of, like,
281 00:33:20.020 ⇒ 00:33:24.039 Katherine Bayless: low-to-no-frills way to interact with, like, Claude and the GitHub repo.
282 00:33:24.040 ⇒ 00:33:35.519 Katherine Bayless: So, like, I basically… I park stuff as, like, issues, and so, I had asked Claude to go through this, like, 3 times, basically. The first pass, I wanted it to look at…
283 00:33:35.600 ⇒ 00:33:47.219 Katherine Bayless: like, some of the, like, standardization stuff. So, like, this repo Jay put together late last fall, I think, but I was like, okay, well, there’s gonna be at least some stuff that’s different
284 00:33:47.220 ⇒ 00:33:57.860 Katherine Bayless: in terms of how his code was written versus, like, the stuff that we’ve written. And so I went through OnePass and was like, okay, come up with all of the issues that we need to standardize between the two sort of approaches.
285 00:33:57.860 ⇒ 00:34:12.279 Katherine Bayless: then also go through and look for bugs, because Jay wrote all of this as, like, a glow-up of a pipeline that is running, but he hasn’t actually put it in production yet. So, like, there’s probably at least a few bugs that could get caught right out the gate.
286 00:34:12.310 ⇒ 00:34:36.979 Katherine Bayless: And then I asked on the third pass for it to go through and sort of… because it is a big beast of an application relative to, you know, a data pipeline, right? Like, it’s ingesting from a few sources, swirling them together, pushing some data back into other platforms, sending out Slack notifications, generating an Excel report. Like, there’s a lot more components here, and so I had it kind of chunk it into some deployment phases.
287 00:34:37.230 ⇒ 00:34:56.020 Katherine Bayless: And then set up one that was like, start here! And so basically what I did when I told Kyle to get started was, like, okay, tell Claude to pull, you know, issue 27, and then kind of take it from there, going through working the issues as if they’re kind of like the tickets slash backlog. So, I don’t know, it seems…
288 00:34:56.020 ⇒ 00:34:58.949 Katherine Bayless: To work kind of nicely, but…
289 00:34:59.060 ⇒ 00:35:03.470 Katherine Bayless: Curious to hear the thoughts of others who do these things more frequently than I do.
290 00:35:06.990 ⇒ 00:35:19.279 Kyle Wandel: I mean, from my perspective, it worked really well. So, it only took, I mean, set up everything pretty seamlessly, and then I pushed my way through the CI infrastructure a little bit here and there, just making some changes, but,
291 00:35:19.390 ⇒ 00:35:20.530 Kyle Wandel: Very easy.
292 00:35:21.910 ⇒ 00:35:22.540 Katherine Bayless: Yeah.
293 00:35:23.380 ⇒ 00:35:23.880 Katherine Bayless: I did.
294 00:35:23.880 ⇒ 00:35:24.450 Kyle Wandel: Tiger team.
295 00:35:24.450 ⇒ 00:35:28.119 Katherine Bayless: I need to do the rest of the, kind of, deep dive and review, but yeah.
296 00:35:28.310 ⇒ 00:35:36.839 Kyle Wandel: I feel like it’s really cool because it’s forcing AI to write it, and then review it, and then write it, and then review it again, again, and again, until it fine-tunes it to what it needs to be.
297 00:35:40.590 ⇒ 00:35:44.199 Awaish Kumar: Okay, so are you writing guidelines for that? For that?
298 00:35:44.200 ⇒ 00:35:44.880 Katherine Bayless: I did.
299 00:35:45.450 ⇒ 00:35:48.049 Awaish Kumar: We do that, like, kind of a standard…
300 00:35:48.900 ⇒ 00:36:13.840 Katherine Bayless: Yes, like, I wrote the ClaudeMD file to say, like, to explain that the issues were being, you know, used as kind of the backlog, and then also to say, like, for each one, like, make sure to tie the issue naming conventions and the PR naming conventions together so that, you know, they kind of fit logically. And to always, when putting in a PR, give it the, like, the summary, the test criteria, and a prompt for a different Claude session.
301 00:36:13.840 ⇒ 00:36:18.130 Katherine Bayless: to then review the PR, and then, of course, a human as well, but, like…
302 00:36:18.130 ⇒ 00:36:20.269 Katherine Bayless: Why not at least get one pass with the robot?
303 00:36:21.600 ⇒ 00:36:37.630 Kyle Wandel: So this was the ploy that Catherine’s talking about. She literally just says, get oriented first to read the two document structures, understand what we want to do, understand what I’m trying to ask you to do, and then start from there. And then, like she said, breaks it down by the different deployment phases.
304 00:36:37.630 ⇒ 00:36:43.840 Kyle Wandel: Foundation, setting up the ingest, transforming,
305 00:36:44.080 ⇒ 00:36:58.420 Kyle Wandel: And so, each one of these, like, deployments have, like, 3 or 4 different issues or different bugs that need to be fixed. So it’s like, this works your way through that, or I guess Cloud works its way through it all the time, and then you offer it any suggestions that if it needs… need to be.
306 00:36:58.730 ⇒ 00:36:59.899 Kyle Wandel: I mean, it was…
307 00:37:00.140 ⇒ 00:37:11.769 Kyle Wandel: Super simple, I think it’s one of those things, like I said, it’s like forcing the AI to do the work, and then reviewing that work, and then making it do it again, and then again, until it has a better iteration of what it’s supposed to do.
308 00:37:13.200 ⇒ 00:37:13.950 Awaish Kumar: Okay.
309 00:37:19.970 ⇒ 00:37:26.309 Kyle Wandel: And then… and I was gonna say, Catherine, I can show very quickly, now that I’m on here, the Streamlit apps?
310 00:37:26.970 ⇒ 00:37:31.290 Kyle Wandel: Or the migration you started doing, or we… and then Kai and I are reviewing?
311 00:37:31.690 ⇒ 00:37:32.340 Katherine Bayless: Yeah.
312 00:37:32.730 ⇒ 00:37:47.320 Katherine Bayless: Yeah, yeah, so I think, while you pull it up, I can’t remember, did I actually already tell you… I know I slacked you, Tom, about this, but, I can’t remember if I told everybody else. So basically, I remembered we had the PBIX files for the Power BI dashboards in an S3 bucket.
313 00:37:47.320 ⇒ 00:37:55.730 Katherine Bayless: And so I was like, okay, Claude, you’ve got the AWS CLI, and the Snowflake CLI, and this, you know, big brain in the middle, so go pull the Power BI files.
314 00:37:55.730 ⇒ 00:38:14.400 Katherine Bayless: figure out which ones we actually need to recreate, because there’s redundancy, there’s also just, like, super stale stuff in there, right? And so create this, like, pipeline step that goes through them holistically and figures out what are the set that we actually need, then go to Snowflake, find the data where it’s modeled currently, and then publish them up as Streamlit apps.
315 00:38:14.400 ⇒ 00:38:24.599 Katherine Bayless: And so it took about 90 minutes, and I migrated all of Power BI into Snowflake. I mean, it looks like ass, don’t get me wrong. They’re the ugliest app that I’ve ever seen.
316 00:38:24.690 ⇒ 00:38:34.319 Katherine Bayless: But that’s an easy thing to fix, right? And I asked Kyle, too, to make sure that all of them were pointed at the right, like, most current, like, models for the data and stuff like that.
317 00:38:34.520 ⇒ 00:38:36.509 Katherine Bayless: But yeah, I’ll let Kyle take it from here.
318 00:38:37.680 ⇒ 00:38:55.269 Kyle Wandel: But yeah, I mean, that’s what we’re starting to do now, so I’m just kind of going through the QAs, both Kai and I went through our initial pass. I probably went through it one more time, just to make sure that the review is the correct review, and then make it pretty much force it, like Catherine said, is it using the correct tables? Because the first kind of pass I went through was more of, like.
319 00:38:55.270 ⇒ 00:39:04.519 Kyle Wandel: structural code, rather than just, referencing, so now I want to go back and make it go reference the correct things, making sure. But yeah, this is just, like, the quick little
320 00:39:05.310 ⇒ 00:39:11.280 Kyle Wandel: Like, overview of, like, obviously not the best looking, but, that’s where Kai will make it better.
321 00:39:11.620 ⇒ 00:39:24.129 Chi Quinn: Yeah, that’s what I was taking a look at, for the past couple of days. Excuse me. Just kind of looking over the visuals, and then I did ask Todd to kind of look over and just kind of see, just based on
322 00:39:24.130 ⇒ 00:39:40.230 Chi Quinn: the, you know, the guidelines for the… make sure it’s, accessibility friendly, make sure it’s user-friendly, and everything, and it gave, like, a great detail, so I have some interesting thoughts. I did add some of the comment… I did add comment to the,
323 00:39:40.310 ⇒ 00:39:46.299 Chi Quinn: in GitHub. But, yeah, I’m very excited to take a look and just revamp it to make it
324 00:39:46.680 ⇒ 00:39:47.940 Chi Quinn: A lot more pretty.
325 00:39:48.090 ⇒ 00:39:53.039 Chi Quinn: I mean, tables are nice, but, you know, let’s make it prettier, so…
326 00:39:53.040 ⇒ 00:39:55.709 Katherine Bayless: Thank you. Yeah.
327 00:39:55.970 ⇒ 00:40:09.340 Katherine Bayless: But yeah, no, I was like, this is, like, super cool. And I think, too, like, so I put it in as the PR, like, the whole pipeline of it, right? Like, the code to go get it, and to think about it, and to push it through, and I’m… I kind of think, you know, probably…
328 00:40:09.450 ⇒ 00:40:27.159 Katherine Bayless: maybe one eventual sort of outcome of this, too, is, like, I know what Kai’s been kind of working on is, like, a concept of, like, a visual design sort of aesthetic that we could put in the repo, and then it becomes, you know, like, anytime somebody asks you to, like, create a Streamlit dashboard, like, send it through this evaluation sort of process.
329 00:40:27.160 ⇒ 00:40:35.340 Katherine Bayless: So I don’t know, like, this might actually give us some interesting initial breadcrumbs to play with some of that, like, look and feel governance type stuff.
330 00:40:37.590 ⇒ 00:40:40.490 Katherine Bayless: And also just death to Power BI.
331 00:40:42.980 ⇒ 00:41:01.939 Katherine Bayless: Like, honestly, part of me was like, if I could turn this into, like, a… I mean, it already is, like, a little, you know, SAM template, but, like, package it up and tell the guys it remembers, like, next time you have a client that signs up for data share, so many associations use Power BI, right? I’m like, here, just have them run this and migrate their entire environment to Snowflake while they cook dinner.
332 00:41:03.480 ⇒ 00:41:04.140 Katherine Bayless: No.
333 00:41:04.140 ⇒ 00:41:05.320 Uttam Kumaran: Would be a dream.
334 00:41:05.720 ⇒ 00:41:06.330 Katherine Bayless: Great.
335 00:41:06.330 ⇒ 00:41:07.840 Uttam Kumaran: Great.
336 00:41:07.840 ⇒ 00:41:08.630 Katherine Bayless: I mean…
337 00:41:08.780 ⇒ 00:41:20.209 Kyle Wandel: companies like Power BI are kind of, in my opinion, screwed. When you have companies like Replit out there, and software companies that can develop apps, like, so freaking quickly, these larger companies can’t, and they just can’t even keep up.
338 00:41:21.880 ⇒ 00:41:37.019 Katherine Bayless: Yeah, actually, interesting, like, side note on that, so Jay had built, it was initially in Streamlit, but he migrated it to Node, like, an app for the grant reviews for the foundation, so, like, you know, he ingests the documents, displays all the stuff, you know, blah blah blah.
339 00:41:37.020 ⇒ 00:41:45.979 Katherine Bayless: But he had put into it, like a bug report thing, and when a bug is reported, it goes through his little dark factory and gets fixed and just gets passed back to the UI.
340 00:41:45.980 ⇒ 00:42:03.540 Katherine Bayless: And so now Steve, the guy that runs the foundation, is over there using, essentially, composable interface software. Just, you know, no big deal. No big deal, right? It’s kind of cool. Jay also built the app for the member lounge reservations that I keep soapboxing about needing to exist.
341 00:42:03.760 ⇒ 00:42:05.240 Uttam Kumaran: Yes, nice.
342 00:42:05.500 ⇒ 00:42:06.510 Katherine Bayless: Kinda wild.
343 00:42:06.810 ⇒ 00:42:07.789 Kyle Wandel: Wait, what is that?
344 00:42:08.000 ⇒ 00:42:09.790 Kyle Wandel: You can reserve microphones now?
345 00:42:10.260 ⇒ 00:42:17.759 Katherine Bayless: So you made an iOS app that shows you the available spaces and time slots, and you can just book it right from your phone.
346 00:42:17.950 ⇒ 00:42:21.890 Katherine Bayless: We’ll eventually loop in the OctaPhees to validate your membership, right?
347 00:42:23.040 ⇒ 00:42:29.849 Kyle Wandel: I mean, like I said, we’re gonna build a better app and website in the next year than we have had in the past 5.
348 00:42:30.210 ⇒ 00:42:30.880 Uttam Kumaran: Yeah.
349 00:42:33.360 ⇒ 00:42:34.620 Katherine Bayless: I’m very excited.
350 00:42:35.050 ⇒ 00:42:36.940 Katherine Bayless: And scared, but mostly excited.
351 00:42:38.330 ⇒ 00:42:58.669 Katherine Bayless: Oh, the other magic trick, that we played with for the first time in Streamlit, slash Snowflake, really, this week, was the, SQL call-out to an LLM. So I did the first pass at that, like, those executive briefers that we have to put together that are like, you know, so-and-so’s having lunch with a person from XYZ company, and they need to know everything they’ve ever done with us.
352 00:42:59.870 ⇒ 00:43:15.990 Katherine Bayless: And so I put this together. Once you search for a company, it takes the snapshot of data returned for the different places in the report and sends it out with a prompt to, Sonnet 4-6, to come back with, like, 3 key insights and 3 key questions.
353 00:43:15.990 ⇒ 00:43:20.759 Katherine Bayless: To ask, just as, you know, some sort of jumping off point. I did actually use Glean.
354 00:43:20.760 ⇒ 00:43:23.089 Katherine Bayless: To draft the language for the prompt.
355 00:43:23.090 ⇒ 00:43:29.880 Katherine Bayless: Because I figured Glean kind of knows us best. And so, did we make changes?
356 00:43:31.270 ⇒ 00:43:35.120 Kyle Wandel: I… I did, but this is… has a new detailed report to it.
357 00:43:36.020 ⇒ 00:43:37.320 Katherine Bayless: Hmm, interesting.
358 00:43:38.350 ⇒ 00:43:44.740 Katherine Bayless: Oh, you know what I mean? It’s also just… no, I think it’s also just that, you’re not in, like, dark mode like I am, so it looks a little better, but…
359 00:43:44.910 ⇒ 00:43:48.520 Katherine Bayless: This does look nicer than mine did. Interesting.
360 00:43:49.360 ⇒ 00:43:51.440 Katherine Bayless: Maybe I pushed changes and forgot.
361 00:43:52.800 ⇒ 00:44:00.250 Kyle Wandel: There you go. But this is the executive brief you’re talking about, and then the… where is the key insights? Where is that? It’s at the bottom?
362 00:44:01.400 ⇒ 00:44:02.350 Katherine Bayless: It would be.
363 00:44:02.350 ⇒ 00:44:23.780 Katherine Bayless: They take a minute to load, actually. Sorry, I should say that. I did notice this, it’s like, it’s, it’s pretty sluggish how long it takes, which could be that I’m overwhelming it with data, that the prompt is clunky, or that the call is slow. Like, I don’t know that I can entirely blame the model just yet. I’m sure that my code could be a little better. But yeah, once it finishes loading, they will appear.
364 00:44:23.780 ⇒ 00:44:24.340 Katherine Bayless: In theory.
365 00:44:24.340 ⇒ 00:44:31.869 Uttam Kumaran: One thing I’ve had success with is I write… I write, and I get, like, a piece of code or a system to work, and then I have, like.
366 00:44:32.460 ⇒ 00:44:40.550 Uttam Kumaran: you know, like, Codex 5.3 or something, do a pass for speed and optimization and testing. Like, kind of a new thread.
367 00:44:40.760 ⇒ 00:44:47.750 Uttam Kumaran: I find, like, doing all of that in one is, like, a little bit tough, so I, like, get it working, and then I’m like, okay, there’s definitely optimizations.
368 00:44:47.860 ⇒ 00:44:51.910 Uttam Kumaran: Find them and, like, add unit tests and things like that, in case that’s, like, helpful.
369 00:44:52.850 ⇒ 00:45:06.459 Katherine Bayless: Yeah, I’ve actually… I’ve kind of started adopting the same pattern, right? It gets at that, like, execution is cheap now, it’s, like, easier for me to throw something together, react to it, and clean it up, versus, like, trying to get it right the first time.
370 00:45:08.030 ⇒ 00:45:08.870 Katherine Bayless: Yeah.
371 00:45:09.190 ⇒ 00:45:21.249 Katherine Bayless: But I did get a chance to show this to Kinsey the other day, and she was suitably intrigued at the ability to, like, make the LLM call within the data right there.
372 00:45:22.400 ⇒ 00:45:41.570 Katherine Bayless: also a good segue to mentioning, I think there’s a lot of interest in getting Snowflake, into potentially Claude, but then Glean, which I know we’ve talked about in the past, because technically Glean is the one enterprise-wide tool that we are using, and so I think maybe…
373 00:45:41.660 ⇒ 00:45:58.660 Katherine Bayless: maybe in the next couple weeks, we can kind of revive that conversation and start really putting some heads around, like, okay, how do we want Snowflake and Glean to interact? Like, what does that infrastructure look like? I mean, can we deliver visuals? Do we want to deliver visuals? All those sorts of questions. I think we could start to kind of dig in on.
374 00:46:00.880 ⇒ 00:46:02.080 Uttam Kumaran: Okay, okay, great.
375 00:46:03.980 ⇒ 00:46:04.590 Awaish Kumar: Okay.
376 00:46:06.280 ⇒ 00:46:11.789 Katherine Bayless: Okay, I know that was, like, a total enormous brain dump. Questions, thoughts, comments, delusions of grandeur?
377 00:46:15.710 ⇒ 00:46:20.679 Uttam Kumaran: I’m just pumped to see, like, all this data landing. That’s awesome. I feel like that was a huge… just a huge win, like…
378 00:46:21.100 ⇒ 00:46:26.359 Uttam Kumaran: I feel like we didn’t think that we were gonna get some of this until a lot later, so this gives us plenty of time.
379 00:46:26.750 ⇒ 00:46:27.460 Katherine Bayless: Yeah.
380 00:46:28.350 ⇒ 00:46:44.640 Katherine Bayless: Yeah, no, I’m actually… I’m very excited to start really seeing some of the different stuff that’s in there. And I also added the… the Snowflake public, like, free data… data share to the account. For the foundation guy, the 990 stuff is in there, so that’s really useful for him with the grant, proposals and whatnot.
381 00:46:44.640 ⇒ 00:46:52.119 Katherine Bayless: But I think Kyle also wants to kind of take a spin in there and look for, like, we do track, like, the entity list and some of those other government data sets.
382 00:46:52.140 ⇒ 00:47:11.309 Katherine Bayless: And then we’ve got this guy in market research who I think could also probably poke around in there and look at, like, you know, are there other things that we could use to add interesting context to some of our analytics? I did ask… I took a stab at, when I was sitting with Kinsey, asking Claude to, like, you know, look at our data and look at the free data and see if there’s anything interesting, and it…
383 00:47:11.580 ⇒ 00:47:17.809 Katherine Bayless: came back with, something about how, this year at CES we had the coldest morning ever, and I was like, well…
384 00:47:18.230 ⇒ 00:47:26.529 Katherine Bayless: That’s a fun fact. That’s nice. I was like, I don’t know that that’s gonna drive a business decision, but I’ll pack a jacket next year.
385 00:47:29.160 ⇒ 00:47:36.990 Katherine Bayless: But yeah, no, I think I really… I feel like the momentum is starting to really kind of build here.
386 00:47:37.150 ⇒ 00:47:42.090 Katherine Bayless: So, I think, I think it’s gonna be… I think it’s gonna be really interesting to see where it kind of goes next.
387 00:47:42.670 ⇒ 00:47:43.220 Uttam Kumaran: Oh.
388 00:47:43.330 ⇒ 00:47:44.270 Uttam Kumaran: Yeah. Great.
389 00:47:45.490 ⇒ 00:47:49.130 Kyle Wandel: Tim’s use them 100% on board, in my opinion, so that means that.
390 00:47:49.130 ⇒ 00:47:50.150 Uttam Kumaran: No way.
391 00:47:50.150 ⇒ 00:48:00.169 Kyle Wandel: we’ll go, and Catherine didn’t really allude to it, but she did a little bit. Even if we do get pushback from some teams, it kind of sounds like
392 00:48:00.320 ⇒ 00:48:02.920 Kyle Wandel: Kinsey wants us to control a lot of it, so…
393 00:48:03.240 ⇒ 00:48:07.510 Kyle Wandel: Depending upon what that… that pushback is, we can go from there, but hopefully not.
394 00:48:07.510 ⇒ 00:48:08.130 Uttam Kumaran: Great.
395 00:48:09.230 ⇒ 00:48:10.430 Uttam Kumaran: That’s a huge win.
396 00:48:11.030 ⇒ 00:48:23.900 Katherine Bayless: It is. And I think I was able to kind of get to some of the, like, the real core of the anxieties around the rollout. Like, I do think she does not think everybody at the organization should have access to it fully, which is…
397 00:48:24.070 ⇒ 00:48:25.469 Katherine Bayless: Correct, right? I mean, it’s.
398 00:48:25.470 ⇒ 00:48:26.130 Uttam Kumaran: Yeah.
399 00:48:26.130 ⇒ 00:48:43.560 Katherine Bayless: yeah, does not really bother me. She didn’t have any sort of, like, only VPs can see it, which that would have been where I’d be like, no, we’re not gonna play that game. But yeah, like, keeping the interns out, the people that know, like, 5 minutes’ worth of information about the business, yeah, that seems fair. But her anxieties were really more around just, like.
400 00:48:43.560 ⇒ 00:48:50.590 Katherine Bayless: are we giving people enough, you know, support to know how to use this well? And I’m like, oh, well, yeah, that’s, like.
401 00:48:50.850 ⇒ 00:49:03.480 Katherine Bayless: kind of what our team is here for, right? Because she also, importantly, you know, and Kyle kind of mentioned this to a certain extent, like, she does not want us to be the only people who analyze or find insights in the data, right? Like, I think, in a weird.
402 00:49:03.480 ⇒ 00:49:03.940 Uttam Kumaran: True.
403 00:49:03.940 ⇒ 00:49:18.429 Katherine Bayless: she’s already where my head is at with our team, which is like, you know, we’re the platform and product team, we build the thing that people can use, and then hopefully we’ll get a chance to do some cool stuff on top of it too, but, like, we’re not the answer factory for the entire organization. And so I think that’s where she’s, like.
404 00:49:18.430 ⇒ 00:49:35.559 Katherine Bayless: no, I don’t want everybody in here, but I do want many, many, many people in here, and I want them to be using it well and frequently. And so I think, like, a lot of the stuff we’ve talked about with, like, the, you know, what questions the data is best suited to ask, and, like, good prompting, you know, techniques that are
405 00:49:35.580 ⇒ 00:49:48.860 Katherine Bayless: not the garden variety, how to prompt AI, but, like, specific to, like, our data in Snowflake and that kind of stuff, so… But yeah, she’s, she’s definitely all in. She wants to tie it to Gamma, by the way, too, and have it be.
406 00:49:48.860 ⇒ 00:49:49.890 Uttam Kumaran: Oh, nice.
407 00:49:50.160 ⇒ 00:49:52.790 Katherine Bayless: the salesperson has a call with Verizon on their…
408 00:49:52.790 ⇒ 00:49:53.900 Uttam Kumaran: Oh, yeah, heck yeah.
409 00:49:53.900 ⇒ 00:49:56.410 Katherine Bayless: Tomorrow, generate a deck for them, yeah.
410 00:49:56.550 ⇒ 00:49:57.430 Uttam Kumaran: Nice.
411 00:49:57.440 ⇒ 00:50:00.609 Katherine Bayless: Yeah, I told Jay to start working on a gamma procurement.
412 00:50:03.100 ⇒ 00:50:13.330 Uttam Kumaran: Very cool. Yeah, we’ve been using a lot of, like, the Google Workspace, like, being able to create slides faster, like Google Slides, and it’s been really, really helpful.
413 00:50:15.530 ⇒ 00:50:23.229 Katherine Bayless: Actually, okay, sorry, this is, like, a small thing, but I kept meaning to ask and forgetting. The Claude in Excel plugin, should we…
414 00:50:24.050 ⇒ 00:50:24.920 Katherine Bayless: Should we look at that?
415 00:50:24.920 ⇒ 00:50:28.350 Uttam Kumaran: Yeah, it’s good. I feel like it’s good.
416 00:50:29.350 ⇒ 00:50:34.630 Uttam Kumaran: I don’t know if it has, like, any ex… if it can really externally reference much, but…
417 00:50:35.140 ⇒ 00:50:42.290 Uttam Kumaran: I used it a couple times to do, like, hey, I need to take this data and, like, reformat it, and it was really, really good.
418 00:50:44.760 ⇒ 00:50:45.120 Katherine Bayless: Because I’ve.
419 00:50:45.120 ⇒ 00:50:56.019 Uttam Kumaran: you need… it’s… you need to kind of, like… I don’t… I think the hardest part for me was actually just, like, how do I get it on the Microsoft Marketplace for our company? That was harder than…
420 00:50:56.210 ⇒ 00:50:59.690 Uttam Kumaran: Like… the whole thing. So…
421 00:50:59.690 ⇒ 00:51:00.180 Katherine Bayless: That tracks.
422 00:51:00.180 ⇒ 00:51:05.409 Uttam Kumaran: If you’re able to figure that out, and then, yeah, it was actually… it’s really helpful for folks.
423 00:51:05.530 ⇒ 00:51:07.540 Uttam Kumaran: Directly in Excel to do work.
424 00:51:08.210 ⇒ 00:51:09.080 Katherine Bayless: Okay.
425 00:51:10.030 ⇒ 00:51:14.260 Kyle Wandel: It might be a good middle ground for not allowing everybody to have Cloud Code, yeah.
426 00:51:16.480 ⇒ 00:51:31.159 Katherine Bayless: That was where my brain was going, is like, is it a, like, a compromise and also a training tool, right? Like, get comfortable using AI on data right here in one Excel sheet where you can’t really break too much and then, you know, kind of graduate folks, so…
427 00:51:31.160 ⇒ 00:51:39.459 Katherine Bayless: But yeah, probably one of us, or all of us, should maybe take a, you know, couple hours, play with it, figure out how to socialize it across the org.
428 00:51:39.460 ⇒ 00:51:55.829 Katherine Bayless: Jay is actively working on figuring out, could we get Claude co-work, right, like, released more broadly? I mean, I think there’s, you know, there’s a lot of questions around the observability and security, and then also very much so the trust component. But I think…
429 00:51:56.060 ⇒ 00:52:05.349 Katherine Bayless: I don’t know. I feel like we will probably be able to roll it out maybe, like, later in the summer, kind of a thing, to a lot of folks at the organization, which would be cool.
430 00:52:05.610 ⇒ 00:52:14.220 Katherine Bayless: We also might go the cohort route, right? Like, you know, take 8 or 10 people, work with them really intensively, and then try to, like, get that to scale out, but…
431 00:52:14.830 ⇒ 00:52:18.160 Katherine Bayless: Yeah, there’s a lot of support right now. A lot of support.
432 00:52:19.960 ⇒ 00:52:20.640 Katherine Bayless: Yeah.
433 00:52:20.980 ⇒ 00:52:21.940 Katherine Bayless: It’s exciting.
434 00:52:24.670 ⇒ 00:52:30.910 Katherine Bayless: Okay, well, the last tiny update, that I had on my brain was, just that the AWS ProServe stuff is…
435 00:52:31.040 ⇒ 00:52:50.970 Katherine Bayless: in progress. I do think I’d like to get you guys into one of the calls and sessions, kind of maybe either next week or the week after. I’ll look at the roadmap again for when we’re getting to what, but because we’ll be migrating, you know, some of the AWS infrastructure around Snowflake, I just figure it makes sense to keep us all in the loop on that one. But yeah.
436 00:52:50.970 ⇒ 00:52:55.120 Katherine Bayless: you know, Jay’s bored to tears in that project, I think. But I’m like.
437 00:52:55.120 ⇒ 00:52:55.580 Uttam Kumaran: Okay.
438 00:52:55.580 ⇒ 00:53:07.959 Katherine Bayless: foundational work, it’s good questions to, like, you know, surface and go through, and it’s good exposure for Kyle and Ian, who haven’t, like, really been through something like that level of a, you know, infrastructure deployment before, so…
439 00:53:09.130 ⇒ 00:53:10.300 Katherine Bayless: Awesome. Bye.
440 00:53:14.720 ⇒ 00:53:16.410 Katherine Bayless: Okay, those are all the things.
441 00:53:18.570 ⇒ 00:53:27.129 Uttam Kumaran: Okay, awesome. I should have some updates today, or probably Monday, on some improvements for Coco, so I’ll send those over.
442 00:53:29.410 ⇒ 00:53:37.580 Uttam Kumaran: But yeah, I feel like this… this also, I think, like, a way she kind of gloss over, but the snowpipe stuff is a huge win, like, it’s… it’s really gonna replace, like.
443 00:53:37.960 ⇒ 00:53:43.880 Uttam Kumaran: a ton of ETL. I did a bunch of snowpipe stuff a few years ago, and it’s, like, a really, really amazing technology.
444 00:53:44.020 ⇒ 00:53:50.469 Uttam Kumaran: Or real-time syncing, so… Like, I think we’re basically gonna be able to…
445 00:53:50.840 ⇒ 00:53:55.850 Uttam Kumaran: I mean, it’s all basically real-time as stuff lands in S3, so it’s really, really good.
446 00:53:56.240 ⇒ 00:53:56.990 Katherine Bayless: Okay.
447 00:53:57.290 ⇒ 00:54:02.979 Katherine Bayless: Yeah, maybe on, Monday, Oish, if you don’t mind, we could do, like, a little demo, or, you know, kind of walkthrough of how it all.
448 00:54:02.980 ⇒ 00:54:03.840 Uttam Kumaran: Yeah.
449 00:54:04.080 ⇒ 00:54:05.039 Awaish Kumar: Yeah. It’s an interesting…
450 00:54:05.040 ⇒ 00:54:23.850 Uttam Kumaran: technology that’s sort of as a listener, and yeah, I think a way should be good for people to see, because it’s a great alternative, and it’s a great alternative for us to continue to use S3, and then quickly move stuff. So I think your point about, like, how do we quickly spin these up for more buckets or more things is best. So, yeah, let’s do that on Monday.
451 00:54:24.400 ⇒ 00:54:25.660 Katherine Bayless: Okay, nice.
452 00:54:25.660 ⇒ 00:54:26.680 Awaish Kumar: Okay, sure.
453 00:54:27.900 ⇒ 00:54:28.540 Katherine Bayless: Like it.
454 00:54:29.690 ⇒ 00:54:34.650 Katherine Bayless: Mmm… Anybody else got stuff on your brains?
455 00:54:35.510 ⇒ 00:54:36.710 Katherine Bayless: Friday thoughts?
456 00:54:42.260 ⇒ 00:54:42.940 Katherine Bayless: Cool.
457 00:54:43.450 ⇒ 00:54:47.919 Uttam Kumaran: Yeah, I guess it might… maybe one thing we can start thinking about in terms of, like.
458 00:54:48.100 ⇒ 00:54:54.710 Uttam Kumaran: kind of post-rollout for Coco is, like, there’s this Snowflake, like, agents technology that I don’t think I really, like.
459 00:54:54.710 ⇒ 00:54:55.850 Katherine Bayless: Yeah.
460 00:54:55.850 ⇒ 00:55:00.070 Uttam Kumaran: in or considered, but… Wondering maybe how we can leverage
461 00:55:00.660 ⇒ 00:55:09.170 Uttam Kumaran: leverage those, so maybe it’s, like, a good brainstorming session for us to refresh on, like, what AI capabilities are out now beyond the sidebar.
462 00:55:10.810 ⇒ 00:55:30.010 Katherine Bayless: I actually… I was curious about that, like, I know it’s kind of silly, but if for no other reason than from an interface perspective, like, if it seems like as long as we have one agent, I could presumably, like, point people here as a first place, and, like, then that’s, you know, a little bit less sort of intimidating.
463 00:55:30.440 ⇒ 00:55:36.890 Katherine Bayless: than the regular SnowSite UI, and so I was curious, like, could we make, like, a super generic agent,
464 00:55:36.890 ⇒ 00:55:40.220 Uttam Kumaran: Yeah, just like a CTA data analyst.
465 00:55:41.000 ⇒ 00:55:41.390 Katherine Bayless: Right?
466 00:55:41.390 ⇒ 00:55:47.070 Uttam Kumaran: Yeah… Okay, great. Yeah, maybe it’s something that we can just try out and see…
467 00:55:47.210 ⇒ 00:55:50.960 Uttam Kumaran: If people enjoy using that, Morning.
468 00:55:50.960 ⇒ 00:55:51.420 Katherine Bayless: Yes.
469 00:55:51.420 ⇒ 00:55:54.089 Uttam Kumaran: It’s gonna leverage a lot of the same work, so…
470 00:55:55.480 ⇒ 00:56:01.669 Katherine Bayless: And I guess, am I correct in guessing that this is probably where snow… snow work will…
471 00:56:01.670 ⇒ 00:56:04.039 Uttam Kumaran: Yes, I’m guessing, yeah.
472 00:56:04.040 ⇒ 00:56:05.060 Katherine Bayless: Okay, okay.
473 00:56:05.060 ⇒ 00:56:09.150 Uttam Kumaran: It’s just so weird, I feel like they’re kind of hodgepodging a bunch of products right now.
474 00:56:09.340 ⇒ 00:56:11.850 Uttam Kumaran: So I also don’t have, like, a super clear…
475 00:56:12.800 ⇒ 00:56:18.379 Uttam Kumaran: perspective on, like, what’s actually gonna land, but either way, they’re gonna build it on semantic views, and…
476 00:56:18.510 ⇒ 00:56:20.809 Uttam Kumaran: All of the underlying tech, so…
477 00:56:21.070 ⇒ 00:56:29.689 Uttam Kumaran: I just think we keep trying different… I’m kind of describing these as, like, surfaces, like, we just keep trying different surfaces, you know, whether it’s us using the CLI, whether it’s…
478 00:56:29.880 ⇒ 00:56:36.380 Uttam Kumaran: People using a sidebar, and then maybe there’s even one level higher, which is this work product, where people don’t even…
479 00:56:37.240 ⇒ 00:56:42.129 Uttam Kumaran: See the data, really, or like, yeah, it’s sort of not even a sidebar, it’s, like, the whole thing, you know?
480 00:56:43.320 ⇒ 00:56:52.249 Katherine Bayless: Well, I think, yeah, like, once we’re comfortable, like, putting something into, like, Slack, where people can just, you know, just talk to it, right? As if it’s your teammate, just ask it a question.
481 00:56:53.850 ⇒ 00:56:55.310 Uttam Kumaran: Exactly, exactly.
482 00:56:56.240 ⇒ 00:56:56.920 Katherine Bayless: Yeah.
483 00:56:57.840 ⇒ 00:56:58.729 Katherine Bayless: Very cool.
484 00:57:00.780 ⇒ 00:57:02.409 Katherine Bayless: Yeah, okay, cool.
485 00:57:02.640 ⇒ 00:57:10.669 Katherine Bayless: Well, in that case, I’ve got… actually, I have a call with the Snowflake guy at 11.30, so I’ll go, maybe I’ll ask him about the snowwork stuff.
486 00:57:11.210 ⇒ 00:57:13.240 Uttam Kumaran: Okay, yeah, please.
487 00:57:13.850 ⇒ 00:57:23.230 Katherine Bayless: Cool. Alright, well, happy Friday, I wish us all delightful adventures, until the weekend comes, which might be around 3 o’clock, as far as I’m concerned.
488 00:57:23.790 ⇒ 00:57:26.820 Uttam Kumaran: Thank you all, appreciate it. Enjoy.
489 00:57:26.820 ⇒ 00:57:27.740 Kyle Wandel: Have a good day.
490 00:57:27.760 ⇒ 00:57:28.300 Chi Quinn: Bye.
491 00:57:28.300 ⇒ 00:57:28.900 Uttam Kumaran: Bye.