Meeting Title: Brainforge x CTA: Weekly! Date: 2026-05-01 Meeting participants: Chris Terry, Awaish Kumar, Amber Lin, Chi Quinn, Katherine Bayless, Ashwini Sharma, Uttam Kumaran
WEBVTT
1 00:01:49.270 ⇒ 00:01:51.539 Awaish Kumar: Hi, Chris. My name.
2 00:01:53.850 ⇒ 00:01:54.930 Chris Terry: Hey, how’s it going?
3 00:01:56.010 ⇒ 00:01:56.860 Awaish Kumar: All good.
4 00:01:58.940 ⇒ 00:01:59.490 Chris Terry: Awesome.
5 00:02:24.670 ⇒ 00:02:25.450 Awaish Kumar: Hello?
6 00:02:25.740 ⇒ 00:02:26.850 Amber Lin: Hi, Bear!
7 00:02:27.780 ⇒ 00:02:28.829 Chris Terry: How’s it going?
8 00:02:29.150 ⇒ 00:02:30.460 Amber Lin: Pretty good.
9 00:02:30.570 ⇒ 00:02:34.229 Amber Lin: It’s a good… I mean, it’s Friday, so I’m happy.
10 00:02:35.420 ⇒ 00:02:36.410 Chris Terry: Always a good one.
11 00:02:43.550 ⇒ 00:02:48.380 Amber Lin: Chris, how has your first week been? It’s your first week, right? Or the end of your first week?
12 00:02:48.960 ⇒ 00:02:56.499 Chris Terry: Second week, second week. It’s going alright. Still learning as much as I can, absorbing it all.
13 00:02:56.930 ⇒ 00:02:57.790 Chris Terry: Enough.
14 00:02:59.060 ⇒ 00:03:01.479 Chris Terry: Same old, we just kind of learn everything, so…
15 00:03:02.560 ⇒ 00:03:03.620 Amber Lin: Hi, Kai.
16 00:03:04.400 ⇒ 00:03:05.630 Chi Quinn: Hello!
17 00:03:39.810 ⇒ 00:03:42.380 Katherine Bayless: Hello! Sorry, guys, change.
18 00:03:42.380 ⇒ 00:03:42.970 Amber Lin: Hi there!
19 00:03:43.780 ⇒ 00:03:54.979 Katherine Bayless: Jay and I were chatting about, an interesting, thing that has come up for us this week. I need, like, 2 more minutes to finish out what I was doing, but I’ll be fully present momentarily.
20 00:03:55.260 ⇒ 00:03:56.820 Amber Lin: Okay, sounds good.
21 00:03:56.820 ⇒ 00:03:57.430 Katherine Bayless: Hell yeah.
22 00:03:58.080 ⇒ 00:04:02.060 Amber Lin: Let’s see…
23 00:04:09.470 ⇒ 00:04:28.180 Amber Lin: I can get started on my side. I can show you guys the… the agent that I’ve been working on. So, I personally think it’s much better than what we had last week. And then I have some agenda items I want to talk about, but other than that, we could use this as a working session if we need to.
24 00:04:30.000 ⇒ 00:04:34.510 Amber Lin: So… Let me log in first.
25 00:04:54.510 ⇒ 00:04:59.329 Amber Lin: Kai and Chris, you both have, like, role developer, right, in Snowflake?
26 00:04:59.540 ⇒ 00:05:00.440 Chi Quinn: Yes.
27 00:05:00.780 ⇒ 00:05:06.750 Amber Lin: Awesome. Okay. So…
28 00:05:11.710 ⇒ 00:05:17.599 Amber Lin: So, I have been working on… the agent, so…
29 00:05:18.120 ⇒ 00:05:26.439 Amber Lin: there’s… I think there’s two ways you can see what’s in it. So, right now, the agent will live under Cortex Agents.
30 00:05:26.570 ⇒ 00:05:44.159 Amber Lin: And then some related artifacts are, under Cortex Search, and then for the semantic views, it will be under semantic analysts. So, if you want to go look at any of the views that’s used, it will be in here, and…
31 00:05:44.760 ⇒ 00:05:48.070 Amber Lin: How you would test this agent is you would go to
32 00:05:48.220 ⇒ 00:05:51.430 Amber Lin: Click on this tab, and click on Agents.
33 00:05:52.610 ⇒ 00:05:57.950 Amber Lin: And… Let’s click on…
34 00:05:58.610 ⇒ 00:06:06.589 Amber Lin: So the one that I’ve been working on, I named it Agent Amber Test Copilot. I would rename this back.
35 00:06:06.710 ⇒ 00:06:22.150 Amber Lin: Before we roll out. And then I also added, if you saw my last message, an agent for last resort that has access to the Broadmars data, because we still have some topics that’s not covered by semantic views.
36 00:06:22.870 ⇒ 00:06:28.170 Amber Lin: So, if we click… In here… Hmm.
37 00:06:29.210 ⇒ 00:06:31.320 Amber Lin: Alright, my computer’s frozen.
38 00:06:35.450 ⇒ 00:06:50.370 Amber Lin: Okay, so if we’re in here, you can see this interface, and this lists the instructions that we have, the different tools it has access to, so it
39 00:06:50.430 ⇒ 00:06:59.100 Amber Lin: Right now, it can’t do web search, but it has access to these analysts, and it has access to Cortex Search.
40 00:06:59.820 ⇒ 00:07:17.670 Amber Lin: And you can test the agent in this interface, but you can also do click preview in Snowflake Intelligence. I recommend doing this to feel what’s closer to the end user, and
41 00:07:18.090 ⇒ 00:07:26.619 Amber Lin: When you’re in here, you can select which one you want to use, and then you can also just click on these.
42 00:07:27.350 ⇒ 00:07:32.309 Amber Lin: these questions. Let’s say… let’s say this.
43 00:07:43.880 ⇒ 00:07:50.279 Amber Lin: So, if you want to see what it’s searching, you can click on this one, and then it will have
44 00:07:50.530 ⇒ 00:07:52.990 Amber Lin: The steps of what it is.
45 00:07:53.330 ⇒ 00:07:58.770 Amber Lin: what it’s doing… Cool.
46 00:07:59.270 ⇒ 00:08:06.799 Amber Lin: I think in this case, because this was a verified query, it went straight ahead to…
47 00:08:07.240 ⇒ 00:08:26.100 Amber Lin: this query. Usually, I ask it to also do a planning step, and if the planning step has some ambiguous questions, it would ask that to us. So once we finish this question, we’ll try one that’s a little bit more, confusing.
48 00:08:27.250 ⇒ 00:08:33.570 Amber Lin: Oh, let’s try… 420… 26…
49 00:08:54.660 ⇒ 00:09:06.289 Amber Lin: And so this one right now has access to the registration and Attendance Analyst, has access to, the CES session and the C event.
50 00:09:06.470 ⇒ 00:09:10.170 Amber Lin: Data, and also exhibitor and membership.
51 00:09:11.150 ⇒ 00:09:21.960 Amber Lin: So… We have this. I… I remember this was quite close… Oh, interesting.
52 00:09:26.210 ⇒ 00:09:40.779 Amber Lin: I remember that was decently close to what we have as the end result. I can go ahead and do some tuning there. But this is what it will respond with, and then you’ll have these questions at the bottom that you can click
53 00:09:41.000 ⇒ 00:09:44.119 Amber Lin: To, dig deeper.
54 00:09:44.490 ⇒ 00:09:46.999 Amber Lin: Let’s try this one.
55 00:09:48.760 ⇒ 00:09:50.499 Amber Lin: And you can always…
56 00:09:51.030 ⇒ 00:10:05.049 Amber Lin: click one of these buttons, and this will pop up to send in the feedback, which I found really helpful, and we’re working on bringing that data in, into our adoption dashboard, or at least into our database.
57 00:10:14.520 ⇒ 00:10:15.989 Amber Lin: And then we have this.
58 00:10:17.000 ⇒ 00:10:28.170 Amber Lin: I think the numbers are a little bit off, but the, like, probably within the 90% or 95% range. So…
59 00:10:28.590 ⇒ 00:10:39.470 Amber Lin: This is how a typical question would get… would get asked. I’m gonna pause here for any questions before I go and try a very ambiguous question to see what it says.
60 00:10:42.030 ⇒ 00:11:00.400 Chi Quinn: So it looks like you have to create, it looks like for the objects, like, for each topic. So, like, one for, sessions, one for, what is it, exhibitors. So that’s the best approach, just to do, like, one, specific topics.
61 00:11:00.940 ⇒ 00:11:05.839 Amber Lin: That’s what I’m doing right now. To be very honest, in membership.
62 00:11:05.890 ⇒ 00:11:13.449 Amber Lin: There’s still membership and attendance. Sometimes you can’t… I can’t avoid, but still having to add some facts
63 00:11:13.490 ⇒ 00:11:28.320 Amber Lin: registration or attendance tables into membership, but overall, this helps the agent knows where to look and how to query, and it’s less likely to generate, confusing SQL or mismatched names.
64 00:11:29.630 ⇒ 00:11:32.589 Awaish Kumar: So yeah, so generally the best practice is that
65 00:11:33.280 ⇒ 00:11:41.570 Awaish Kumar: The semantic view that we are creating is… Is… Specific to some domain.
66 00:11:42.430 ⇒ 00:11:48.519 Awaish Kumar: A narrow domain or topic that we want to get answers for, and…
67 00:11:48.630 ⇒ 00:11:53.550 Awaish Kumar: Tune it to answer best possible results for that specific domain.
68 00:11:53.770 ⇒ 00:11:56.600 Awaish Kumar: And, and we create, maybe.
69 00:11:56.740 ⇒ 00:12:03.399 Awaish Kumar: Multiple of these semantic views, and then put them all together in a single agent.
70 00:12:03.610 ⇒ 00:12:08.760 Awaish Kumar: So, people can still get the answers, but we have more control.
71 00:12:09.150 ⇒ 00:12:12.350 Awaish Kumar: overtuning and… And preparing it.
72 00:12:14.140 ⇒ 00:12:26.020 Chi Quinn: Oh, okay, that makes sense. That makes sense. So, and you could put as many in one agent, and so it would just be helpful for the agent to specify, based on the question, it would go to that topic.
73 00:12:26.980 ⇒ 00:12:38.370 Amber Lin: Yeah, I can show you how I did the routing, and guidelines in a bit, but this is one of the results I was working on from yesterday. So, sometimes we have very…
74 00:12:38.720 ⇒ 00:12:51.049 Amber Lin: sometimes the user might not specify it very clearly, or they might not even know what they want, and I don’t want the agent to go down a rabbit hole when they don’t exactly know what they’re being asked.
75 00:12:51.060 ⇒ 00:12:59.859 Amber Lin: So in this case, we asked something, program reach for innovation. We might know this is for the Innovation Awards, but for the agent.
76 00:13:00.430 ⇒ 00:13:10.170 Amber Lin: innovation might mean different things. It might mean a product of interest, it might mean a… it might mean a session. So, in that case, I’ve asked it to
77 00:13:10.470 ⇒ 00:13:28.099 Amber Lin: clarify, what it means. It says I should not run the analyst, I should go with clarify. So, it will ask us, hey, these are the possible interpretations, what are you really looking for? I’m not going to execute a query until…
78 00:13:28.190 ⇒ 00:13:38.570 Amber Lin: you tell me. So, in this case, we can say it’s with… This one, and…
79 00:13:41.380 ⇒ 00:13:43.860 Amber Lin: And we’ll go back and think again.
80 00:13:44.370 ⇒ 00:13:48.239 Amber Lin: This is a question that’s also outside
81 00:13:49.210 ⇒ 00:13:57.540 Amber Lin: of our semantic model. So, in this case, we have the planning next steps, and it’ll say, hey, we don’t have anything
82 00:13:57.750 ⇒ 00:14:00.670 Amber Lin: In the semantic fuse.
83 00:14:00.870 ⇒ 00:14:02.110 Amber Lin: And…
84 00:14:04.640 ⇒ 00:14:11.529 Amber Lin: It’ll say, hey, we don’t have these, I can’t find any of these. So in this case, we can go to…
85 00:14:11.830 ⇒ 00:14:14.770 Amber Lin: We can change to this one.
86 00:14:15.030 ⇒ 00:14:17.499 Amber Lin: And then we can go ask.
87 00:14:18.080 ⇒ 00:14:22.870 Amber Lin: Let’s say try again with this agent.
88 00:14:42.450 ⇒ 00:14:43.160 Amber Lin: Oh.
89 00:14:43.430 ⇒ 00:14:47.390 Amber Lin: I don’t think I understood this. This was the Innovation Awards.
90 00:14:47.730 ⇒ 00:14:57.920 Amber Lin: Innovation Awards… Attendance?
91 00:14:58.530 ⇒ 00:15:01.569 Amber Lin: Or count, I’m not very sure of.
92 00:15:01.720 ⇒ 00:15:08.719 Amber Lin: what I should be asking, but I at least know if it doesn’t know, it will ask me. So…
93 00:15:08.850 ⇒ 00:15:10.870 Amber Lin: Let’s see what it comes up with.
94 00:15:29.830 ⇒ 00:15:31.320 Amber Lin: Okay.
95 00:15:31.500 ⇒ 00:15:37.000 Amber Lin: I think it got… I think it got confused again.
96 00:15:37.380 ⇒ 00:15:45.000 Amber Lin: But anyways, I’m gonna pause here to not take up too much time, but this is… this is how I wanted to…
97 00:15:45.530 ⇒ 00:16:03.720 Amber Lin: answer a question has not dive too deep on a specific thing, because I remember last time, it went… it spent a lot of time trying to reason itself within the pre-audit doc, when the user should know what they’re asking about, and…
98 00:16:03.790 ⇒ 00:16:13.039 Amber Lin: if they don’t know, we should at least have clear guidance of, hey, here is where you can look up things. So…
99 00:16:13.350 ⇒ 00:16:16.740 Amber Lin: That’s how we’re… Let me…
100 00:16:17.760 ⇒ 00:16:25.970 Amber Lin: Let’s see… that’s how we’re doing it inside this agent right now. I do want to show you how I did…
101 00:16:26.160 ⇒ 00:16:33.669 Amber Lin: give the agent instructions, so I’m gonna pause here for questions before I move on there.
102 00:16:43.430 ⇒ 00:16:49.099 Amber Lin: Okay, no questions. Let’s… Go over…
103 00:17:11.040 ⇒ 00:17:29.810 Amber Lin: So, I am now in Kirscher, so let me show you… I think Away showed this a little bit, of how we made the first agent, so we made the cortex search, and we constructed the semantic views, and lastly, we combined it together
104 00:17:30.180 ⇒ 00:17:41.009 Amber Lin: in… with instructions, some sample questions, and then the different tools, which are the Cortex Analyst, or the…
105 00:17:41.410 ⇒ 00:17:46.499 Amber Lin: Or the Cortex Search Service.
106 00:17:47.160 ⇒ 00:18:07.009 Amber Lin: So, the instructions for the agent mostly lives in here, and it has system, has orchestration, and response. They’re slightly different, and ideally, we want them to be less overlapping. Let’s take a look at…
107 00:18:07.740 ⇒ 00:18:10.460 Amber Lin: So, under my fall here…
108 00:18:10.940 ⇒ 00:18:16.130 Amber Lin: I made it a markdown file, just so that it’s a bit easier to read.
109 00:18:16.920 ⇒ 00:18:20.020 Amber Lin: So these are the different tools that we have.
110 00:18:20.160 ⇒ 00:18:27.419 Amber Lin: for… For this agent. And then under instructions, we have different systems.
111 00:18:27.710 ⇒ 00:18:34.050 Amber Lin: So, this is telling the agent who you are and how you should
112 00:18:34.180 ⇒ 00:18:51.100 Amber Lin: How you should respond. So, for example, how you should deal with ambiguity is, I put it here, is, hey, don’t use the analyst if you… you are not sure of, something that might have different interpretations.
113 00:18:52.400 ⇒ 00:19:08.290 Amber Lin: And orchestration is more on routing. So, Kai, to your question, this is how it understands what question goes into, CES sessions, or what questions go into the C event, mart.
114 00:19:08.970 ⇒ 00:19:09.510 Chi Quinn: I see.
115 00:19:09.510 ⇒ 00:19:28.909 Amber Lin: And yeah, and then we can specify here, hey, these are specific words you should look out for, this is sometimes where sample questions might be helpful, so I try to put in sample questions where it might be ambiguous and give it examples of how it should route it. And also here.
116 00:19:28.970 ⇒ 00:19:37.690 Amber Lin: Some… a bit more detailed of, hey, how you should exactly behave when you have, ambiguity and follow-ups.
117 00:19:38.510 ⇒ 00:19:50.190 Amber Lin: And lastly, this is… response is about how they pres… how the agent presents its answers. Do we want it to have tables instead of text? Do we want it to have…
118 00:19:50.340 ⇒ 00:19:54.870 Amber Lin: Visualizations. Do we want to have bullets?
119 00:19:55.030 ⇒ 00:20:10.879 Amber Lin: or want to have follow-ups. So these are areas that you can customize, and we can tweak these things as we go, as we get more feedback through people, and then we can iterate on this prompt.
120 00:20:13.580 ⇒ 00:20:17.330 Amber Lin: Cool. And then… I’m gonna pause here.
121 00:20:17.540 ⇒ 00:20:19.220 Amber Lin: Any questions?
122 00:20:21.120 ⇒ 00:20:22.320 Amber Lin: Comments?
123 00:20:23.550 ⇒ 00:20:24.160 Chris Terry: Yeah, hello.
124 00:20:25.490 ⇒ 00:20:36.400 Chris Terry: I have a quick question. Is all that, through, like, the… is that all being done through GitHub, like, or is that all, like, the Cortex agent and all that stuff being done?
125 00:20:37.650 ⇒ 00:20:45.850 Amber Lin: We are connected to GitHub, so we use it for versioning, and, you should be able… like, all the code should be…
126 00:20:45.850 ⇒ 00:20:46.310 Awaish Kumar: Excuse me.
127 00:20:46.310 ⇒ 00:20:52.679 Amber Lin: in GitHub, but you can create it in the UI, you can create it in the Snowflake UI,
128 00:20:53.720 ⇒ 00:20:56.180 Amber Lin: Yeah, what I’m doing… go ahead, sorry.
129 00:20:56.670 ⇒ 00:21:08.729 Awaish Kumar: Yeah, I would recommend using GitHub. We already… so the one that Amber was showing, she already have it in a GitHub, so you can go and you can look at those,
130 00:21:08.890 ⇒ 00:21:12.650 Awaish Kumar: Files, and it should be the similar approach.
131 00:21:12.930 ⇒ 00:21:18.680 Awaish Kumar: We’ve… the commands and everything will be the exactly same as they are there.
132 00:21:18.850 ⇒ 00:21:23.779 Awaish Kumar: Only thing that will be different is the semantic view.
133 00:21:24.040 ⇒ 00:21:27.429 Awaish Kumar: And, such surveys that we are using.
134 00:21:27.590 ⇒ 00:21:28.940 Awaish Kumar: So, the exact…
135 00:21:29.100 ⇒ 00:21:45.209 Awaish Kumar: tables and columns that you need for any other domain, for any other dataset. We need to update that, and if you have, like, any documents, where to read from. So we have to change those parts and things.
136 00:21:45.540 ⇒ 00:21:50.850 Awaish Kumar: And, but apart from that, the full structure is there, all the steps are there.
137 00:21:51.060 ⇒ 00:21:52.230 Awaish Kumar: And,
138 00:21:53.060 ⇒ 00:22:06.680 Awaish Kumar: basically, and yeah, all the commands are there in that Cortex agent. So, by doing that, it makes it easy, since we use Cortex code or something like that, so it’s like…
139 00:22:06.860 ⇒ 00:22:18.499 Awaish Kumar: You can use AI to basically help you with building those agents, commands, semantic views, and using Sino CLI, we can execute directly from there, and create it in Snowflake.
140 00:22:19.040 ⇒ 00:22:24.280 Awaish Kumar: So, yeah, going into this conflict and clicking around, we can actually…
141 00:22:24.750 ⇒ 00:22:27.000 Awaish Kumar: do that very much in from code.
142 00:22:28.720 ⇒ 00:22:30.499 Chris Terry: Awesome, awesome. Appreciate it.
143 00:22:33.730 ⇒ 00:22:34.330 Amber Lin: Cool.
144 00:22:34.690 ⇒ 00:22:45.000 Amber Lin: So that’s mainly on the agent side. I had a few items on my list, but I want to check if you guys have, any items for doing that you want to talk about.
145 00:22:49.970 ⇒ 00:22:56.829 Katherine Bayless: I was hoping we could kind of do a dive on the exhibitor data, the expoCAD stuff.
146 00:22:56.950 ⇒ 00:22:58.400 Katherine Bayless: Because I…
147 00:22:58.460 ⇒ 00:23:09.530 Katherine Bayless: either I’m doing it wrong, or somewhere we’ve totally gotten a join wrong, because I, rebecca from the international team had asked me to give her a list, or excuse me, a count of
148 00:23:09.530 ⇒ 00:23:34.449 Katherine Bayless: exhibitors by country for Europe and Asia, and the initial list… admittedly, I should have caught it at first. But, like, I guess Kyle was having trouble with it, the numbers were, like, too big. Then I went in and I was like, ahahaha, I think he simply forgot this other joint here. And so I added that, but then they got too small, and then I was… didn’t catch that, and she sent back, like, there should definitely be more for Italy. And I was able… since I still have the raw data, right?
149 00:23:34.450 ⇒ 00:23:38.989 Katherine Bayless: that we, had in the old data warehouse, and so I went back to that, and…
150 00:23:39.150 ⇒ 00:23:53.140 Katherine Bayless: Yeah, the snowflake stuff is totally garbage, and I just can’t, like, I haven’t really had a chance to dig in and figure out, like, okay, where’s the missing piece, but the two things I think I’m running into are the country stuff is a mess, but I…
151 00:23:53.140 ⇒ 00:24:09.019 Katherine Bayless: I mean that, like, the ExpoCAD, and all of our data sources. Like, none of our systems use ISO codes, and so we’re trying to kind of clean up the countries all over the place, and this is one of the spots where we have some of that, like, canonical stuff, but then there’s join keys.
152 00:24:09.020 ⇒ 00:24:15.179 Katherine Bayless: and country names, and so I think the country stuff’s getting a little squirrely. And then the other piece was, like.
153 00:24:15.220 ⇒ 00:24:23.409 Katherine Bayless: child exhibitors seemed to join, but booths did not. And I think the financials are…
154 00:24:23.470 ⇒ 00:24:30.939 Katherine Bayless: maybe causing trouble? So, like, let me actually… let me share my screen.
155 00:24:33.520 ⇒ 00:24:37.780 Katherine Bayless: Yeah, okay, so this one was an example…
156 00:24:42.560 ⇒ 00:24:43.510 Katherine Bayless: Okay.
157 00:24:44.320 ⇒ 00:24:50.600 Katherine Bayless: So this company, Evo Cargo, so I was… United Arab Emirates had,
158 00:24:50.780 ⇒ 00:25:04.719 Katherine Bayless: two exhibitors in the, like, raw, original source data, excuse me, had one, but then I had two over here, and so I was able to validate dubdev is on both, EvoCargo was the outlier. And so…
159 00:25:04.880 ⇒ 00:25:09.460 Katherine Bayless: I asked the little robot, to do some digging,
160 00:25:09.540 ⇒ 00:25:23.840 Katherine Bayless: obviously I can’t see, like, the DPT part of it, but it caught that, like, CES 2022 had an actual, like, booth rental amount, but then the rest of the years were just kind of, like, offsets, or null amounts, and so I’m like.
161 00:25:23.840 ⇒ 00:25:39.690 Katherine Bayless: If I had to guess, this was probably somebody or company that exhibited in 2022, and somehow the way the joins are is, like, making it look like they were exhibiting each year, but I don’t see them anywhere in the raw data for 2026 or 2025.
162 00:25:39.690 ⇒ 00:25:42.889 Katherine Bayless: Which were the two years that I checked, so…
163 00:25:43.030 ⇒ 00:25:56.490 Katherine Bayless: Yeah, I just wasn’t… I wasn’t entirely sure what the kind of best way to tackle this one was, but I figure if we can get to the root cause of where the joins are getting squirrely, we can get them all patched up.
164 00:25:57.760 ⇒ 00:26:02.970 Awaish Kumar: Yeah, like, if you can send that, Carrie, and…
165 00:26:03.640 ⇒ 00:26:08.139 Awaish Kumar: Yeah, so I… yeah, I can… Get it, get debugged.
166 00:26:08.700 ⇒ 00:26:12.130 Awaish Kumar: curating demand fact table from RAW,
167 00:26:12.290 ⇒ 00:26:15.740 Awaish Kumar: There might be some… something happening between that.
168 00:26:16.560 ⇒ 00:26:17.480 Katherine Bayless: Yeah.
169 00:26:17.820 ⇒ 00:26:19.300 Katherine Bayless: And I wonder, so, like.
170 00:26:22.030 ⇒ 00:26:32.510 Katherine Bayless: I don’t know, like, I know I was the person that was like, I want to do the kind of star schema approach, and I’m pretty sure I stand by it, but I am genuinely curious, like, Awish, to get your thoughts on, like.
171 00:26:33.310 ⇒ 00:26:38.409 Katherine Bayless: Am I… like, are we introducing more complexity than, like.
172 00:26:38.530 ⇒ 00:26:52.709 Katherine Bayless: value with it, because I’m like, at the end of the day, these were just flat files, and then it was my idea to be like, break them apart and make them a star schema, because I think it’ll be more performant and better for, like, the different types of analysis we want to do, but now I’m like.
173 00:26:53.650 ⇒ 00:26:57.109 Katherine Bayless: No, maybe I made more complexity than was necessary.
174 00:27:03.640 ⇒ 00:27:08.080 Awaish Kumar: I, I, I don’t, like, I think, like, the…
175 00:27:08.290 ⇒ 00:27:10.750 Awaish Kumar: The way we created those models.
176 00:27:11.450 ⇒ 00:27:21.259 Awaish Kumar: like, is a good approach. The only thing is that validations. So, like, whenever we are… we try to split the…
177 00:27:21.580 ⇒ 00:27:25.010 Awaish Kumar: Table… one single table into multiples, and
178 00:27:25.140 ⇒ 00:27:35.699 Awaish Kumar: Then we created, like, I want to have a DIM table which is really at a company level, and that single company can be across multiple ERs and things like that.
179 00:27:35.850 ⇒ 00:27:39.619 Awaish Kumar: So… By doing that, like, there…
180 00:27:40.740 ⇒ 00:27:54.829 Awaish Kumar: might be some issue which can arrive. So, the main part is the validation at the time when we are building that. So, that step, like, if we maybe do more… we could put more effort on that, like, since we…
181 00:27:54.950 ⇒ 00:27:59.519 Awaish Kumar: Sometimes the, like, from our perspective.
182 00:27:59.700 ⇒ 00:28:02.420 Awaish Kumar: I, like, I know what is in the data.
183 00:28:03.670 ⇒ 00:28:19.579 Awaish Kumar: only, so we don’t have any business domain, like, at least the business domain knowledge. Yeah. Some of the data, like, okay, so what’s going on? Even if I see a few numbers, here and there, like, I’m not able to
184 00:28:19.700 ⇒ 00:28:23.230 Awaish Kumar: judge, like, whether it’s correct or incorrect.
185 00:28:23.650 ⇒ 00:28:24.510 Katherine Bayless: Yeah.
186 00:28:24.920 ⇒ 00:28:25.700 Katherine Bayless: Jump in it.
187 00:28:25.700 ⇒ 00:28:32.640 Amber Lin: Is the number in the pre-auto report correct? I can give you the query that I used to get to that.
188 00:28:33.200 ⇒ 00:28:45.589 Katherine Bayless: It better be, I hope it is. I’m sure it is. So the pre-audit report we did do off of these old flat files, like, we didn’t do off of the dbt model data, but yeah, worth taking a look, actually.
189 00:28:45.590 ⇒ 00:28:48.499 Amber Lin: Okay, I’m including… so when I did it, I…
190 00:28:49.900 ⇒ 00:28:55.230 Amber Lin: Part of this was, I asked the robot to generate a lot of different queries, and.
191 00:28:55.940 ⇒ 00:29:05.100 Amber Lin: get… give me the closest one. I did have to join in the child exhibitors and have to join in the booths, so maybe there’s some…
192 00:29:05.340 ⇒ 00:29:18.849 Amber Lin: I also did a union, so maybe some there, caused a difference, but overall, it gave me numbers that almost exactly matched, so maybe, like, maybe we can learn something off of there.
193 00:29:19.020 ⇒ 00:29:20.990 Katherine Bayless: Yeah, let me rerun this.
194 00:29:20.990 ⇒ 00:29:23.499 Awaish Kumar: For the CS audit report, like…
195 00:29:23.840 ⇒ 00:29:29.189 Awaish Kumar: We mentioned that, like, these were created based on some of the queries that,
196 00:29:29.750 ⇒ 00:29:36.970 Awaish Kumar: Like, the pre-order report that… It was created before we actually created the models in Snowflake.
197 00:29:37.370 ⇒ 00:29:38.790 Katherine Bayless: Yeah, exactly, exactly.
198 00:29:38.790 ⇒ 00:29:46.670 Awaish Kumar: So the data set is also, like, we got a new file also from… GES2026.
199 00:29:48.760 ⇒ 00:29:52.320 Katherine Bayless: Well, yeah, so they’re…
200 00:29:52.880 ⇒ 00:29:57.710 Awaish Kumar: So there was an old file where we didn’t have the product codes, and it was some missing…
201 00:29:58.230 ⇒ 00:30:00.650 Awaish Kumar: Product… like, the product codes and…
202 00:30:00.650 ⇒ 00:30:12.709 Katherine Bayless: Oh, yeah, so that’s, yeah, that’s the other file, but you are right, there was, I think, Dave Hennessy had flagged that issue earlier this week around the product codes, because
203 00:30:12.890 ⇒ 00:30:37.710 Katherine Bayless: That was… that was actually something on the attendee side that came up during the pre-audit report preparation was my counts for, like, how many people were interested in AI were not the same as what Merits was getting on the audit side, and I was like, I don’t know, I was like, this… we’re running the same SQL query, and they realized that the column that has the text values for them, so artificial intelligence, comma, compute.
204 00:30:37.710 ⇒ 00:30:39.940 Katherine Bayless: Comma, automotive, whatever.
205 00:30:40.440 ⇒ 00:31:02.260 Katherine Bayless: gets truncated, because the string is too long for the database, and so they added the column with the actual codes, so that we would be sure we were getting all of the things people were interested in, and then after that, everything matched up. But it does mean that probably the text field is still unreliable, so we… that’s where we had created, like, the lookup dictionary,
206 00:31:02.380 ⇒ 00:31:16.400 Katherine Bayless: Which is actually a small side note, we have, again, changed all the codes, this year, so we’ll need to add another, set of, what do we call them? Equivalencies, to that table, so, like, we… yeah, exactly.
207 00:31:16.400 ⇒ 00:31:36.329 Katherine Bayless: But then, so, but then the exhibitor data, though, this is coming out of the separate… it would have been in, I think it was called, like, the CES26… 2026 exhibitor data set, or something like that, so not the same as the file with the attendees, but this was just the exhibitor file, which is the one that ties to the ExpoCAD Marts.
208 00:31:36.490 ⇒ 00:31:36.980 Katherine Bayless: Yep.
209 00:31:36.980 ⇒ 00:31:46.820 Awaish Kumar: But Amber was saying that, like, the mismatch in some of the numbers, I’m just saying that might reduce the result from that,
210 00:31:47.050 ⇒ 00:31:49.340 Awaish Kumar: Dataset being different, and…
211 00:31:50.240 ⇒ 00:31:52.309 Katherine Bayless: But I mean, like, this one didn’t change.
212 00:31:54.790 ⇒ 00:32:00.200 Awaish Kumar: So for… The audit report, you’re saying, like, no…
213 00:32:02.620 ⇒ 00:32:08.079 Katherine Bayless: Yeah, so for the exhibitors, there wasn’t any change on that data file, like, during the audit.
214 00:32:09.430 ⇒ 00:32:13.289 Awaish Kumar: No, no, for the exhibitors, it’s, it’s, it’s fine.
215 00:32:13.430 ⇒ 00:32:22.879 Awaish Kumar: She’s… I think what she raised about is more about CES data, like, while creating this agent, she’s getting this attendee registrations number and the…
216 00:32:23.720 ⇒ 00:32:24.150 Katherine Bayless: Oh.
217 00:32:24.420 ⇒ 00:32:32.339 Awaish Kumar: Byronage type, and then there is, like, few numbers hidden there, like, you have maybe 10 more attendees and things like that.
218 00:32:33.630 ⇒ 00:32:41.480 Katherine Bayless: Yeah, no, the… the agent, I think, does both. And, like, the query that’s in the chat goes off the ExpoCAD stuff, but it’s not…
219 00:32:44.540 ⇒ 00:33:04.639 Katherine Bayless: it’s not the same as how that… like, the ExpoCAD data hasn’t changed at all. We’ve just changed how we were ingesting it, because before we took the flat file, now we have the actual, like, the raw data and the marts. And so what’s in the agent’s understanding corresponds to, like, the query, like, ampersand against the ExpoCAD data.
220 00:33:04.640 ⇒ 00:33:06.480 Katherine Bayless: Not the attendee data.
221 00:33:07.410 ⇒ 00:33:16.570 Katherine Bayless: Or I might be totally lost, like, correct me if I’m, like, super wrong, but I’m like, this query, I think, is against the ExpoCAD Marts, because it’s part of the exhibitor side of the audit.
222 00:33:18.290 ⇒ 00:33:22.859 Awaish Kumar: I’m not sure, like, if I’m worried it’s just…
223 00:33:22.860 ⇒ 00:33:25.600 Amber Lin: Let’s see, I think… well, let’s…
224 00:33:25.780 ⇒ 00:33:30.209 Awaish Kumar: And are you just talking about, like, the CES data only, or the exhibitor data?
225 00:33:32.340 ⇒ 00:33:40.670 Amber Lin: Well, it should all be in the ExpoCAD. If you look at where the joins are from, it’s all ExpoCAD.
226 00:33:40.780 ⇒ 00:33:48.240 Amber Lin: It’s, like, SVOCADIM Events, SFOCAD DIM countries, SFOCAD…
227 00:33:48.630 ⇒ 00:33:53.669 Amber Lin: gym exhibitor, and then the fact booths assignments and stuff.
228 00:33:54.040 ⇒ 00:33:54.530 Awaish Kumar: Okay.
229 00:33:55.810 ⇒ 00:34:07.029 Katherine Bayless: I mean, unless… I guess the… yeah, the place where I could be getting confused is unless these ExpoCAD marts contain data from not ExpoCAD, in which case, we would just want to fix that.
230 00:34:09.080 ⇒ 00:34:13.300 Katherine Bayless: But I don’t think they do. I think the ExpoCAD marts are only the ExpoCAD data, right?
231 00:34:13.530 ⇒ 00:34:14.120 Awaish Kumar: Yes, yes.
232 00:34:14.120 ⇒ 00:34:16.229 Katherine Bayless: Yeah, okay, okay, okay, okay, gotcha, gotcha.
233 00:34:16.230 ⇒ 00:34:17.899 Awaish Kumar: Can I look into that?
234 00:34:19.179 ⇒ 00:34:21.369 Katherine Bayless: Okay, let’s see…
235 00:34:25.279 ⇒ 00:34:29.599 Katherine Bayless: Why am I getting a funky syntax error on this query?
236 00:34:36.649 ⇒ 00:34:37.889 Katherine Bayless: Okay, now it’s running.
237 00:34:39.159 ⇒ 00:34:48.159 Katherine Bayless: Okay, so… Let’s see, and this is for CES 2020…
238 00:34:48.969 ⇒ 00:34:56.109 Katherine Bayless: 6… okay, so… $960 for the pre-audit, which… let me look at my…
239 00:35:00.959 ⇒ 00:35:03.919 Katherine Bayless: Oh, that’s because I’m looking in the wrong place. Too many things.
240 00:35:19.099 ⇒ 00:35:23.099 Katherine Bayless: Yeah, yeah, so these counts are only different
241 00:35:23.549 ⇒ 00:35:32.909 Katherine Bayless: By 1, which is… I’m… that is perfectly fine by me. So yeah, your logic is probably correct. So then…
242 00:35:32.910 ⇒ 00:35:35.410 Amber Lin: Probably the child exhibitors.
243 00:35:35.720 ⇒ 00:35:36.260 Katherine Bayless: internet.
244 00:35:36.260 ⇒ 00:35:54.849 Amber Lin: union in. The other result I got without that had, like, the first two was okay, they were, like, 100, 200 off, but then everything, like, for example, South Korea was, like, 100-something, so I do think…
245 00:35:54.960 ⇒ 00:35:57.669 Amber Lin: The child exhibitors do matter.
246 00:35:57.860 ⇒ 00:36:03.769 Amber Lin: Yeah, and then for the other countries, they’re also, like, 100 or 50 lower.
247 00:36:05.060 ⇒ 00:36:10.900 Katherine Bayless: Yeah. No, yeah, that does make sense. I wonder… let me see if I try searching on…
248 00:36:11.930 ⇒ 00:36:17.900 Katherine Bayless: That one, company, where did it go? Evo Cargo. There we go.
249 00:36:46.680 ⇒ 00:36:48.260 Katherine Bayless: Hmm, interesting.
250 00:36:54.100 ⇒ 00:36:54.750 Amber Lin: Hi, what’s on?
251 00:36:54.750 ⇒ 00:36:55.460 Uttam Kumaran: Ace.
252 00:36:55.870 ⇒ 00:36:58.220 Amber Lin: We’re doing a working session right now.
253 00:36:58.220 ⇒ 00:36:58.770 Uttam Kumaran: Cool.
254 00:37:01.530 ⇒ 00:37:09.579 Katherine Bayless: Yeah, no, so you’re… it’s funny, yeah, like, your queries… Coming back correctly, it…
255 00:37:10.310 ⇒ 00:37:14.039 Katherine Bayless: Seems like… and, like, in mine, I think…
256 00:37:18.000 ⇒ 00:37:37.280 Katherine Bayless: I think my challenge is probably the financials piece coming in then, because that’s the only real difference, because you’re totally right about, like, unioning the child exhibitors, but I get the right results for the UAE and for that company with your joins, but then when I include the financial information is where I get…
257 00:37:37.320 ⇒ 00:37:38.390 Katherine Bayless: Kinda…
258 00:37:38.660 ⇒ 00:37:43.980 Katherine Bayless: that’s where it starts to get squirrely. And I’d only… it’s funny, I’d only use the financial information, because I just wasn’t…
259 00:37:44.100 ⇒ 00:37:47.269 Katherine Bayless: Sure, if we had unpaid exhibitors?
260 00:37:47.270 ⇒ 00:38:06.180 Katherine Bayless: still in the data, and I was too lazy to just go find that answer out, and so I was like, I don’t know, just join it to the financials, and, you know, filter run isPaid equal true. So I might have just kind of stumbled across, the thing. So, I can do a little bit more digging then, and maybe just, like, if I can pin down where in the financials, we might need to either change the way I’m joining to it, or
261 00:38:06.180 ⇒ 00:38:23.369 Katherine Bayless: maybe that logic is different. To be totally honest, I’ve never worked with the financial data out of ExpoCAD before, so it’s entirely possible that there’s some funky nuance to the way it’s supposed to come together. But that makes more sense than if that’s the missing link, probably, between getting good aggregations and not, so…
262 00:38:23.570 ⇒ 00:38:25.179 Katherine Bayless: Mystery, partly solved.
263 00:38:25.560 ⇒ 00:38:27.710 Amber Lin: Okay, awesome.
264 00:38:27.880 ⇒ 00:38:33.809 Awaish Kumar: that, so, are you getting the low numbers with that data, or higher? Like…
265 00:38:34.280 ⇒ 00:38:40.940 Katherine Bayless: Yeah, no, I get the… with Amber’s query, I get the correct numbers, or, you know, within one or two, which is fine enough.
266 00:38:40.940 ⇒ 00:38:57.390 Katherine Bayless: But it’s the… yeah, like, in my query, the only difference I can see is the financial data’s in there, and so I’ll dig around and see if that’s where either I’m… I might be joining to it wrong, or we might not understand how it is supposed to be set up, because we’ve… I’ve just never worked with the ExpoCAD financial model before.
267 00:38:57.800 ⇒ 00:39:05.310 Awaish Kumar: Yeah, I’m just trying to understand, like, when we join a financial data, like, is it affecting… how it is affecting the numbers, like…
268 00:39:05.760 ⇒ 00:39:09.629 Katherine Bayless: Oh, when I join the financial data, the numbers get too big.
269 00:39:10.900 ⇒ 00:39:19.580 Awaish Kumar: Okay, so, okay, there might be… Duplications or more entries in… Financial data, alright.
270 00:39:19.580 ⇒ 00:39:29.349 Katherine Bayless: Right, well, and I think that’s why I’m like, that data is probably really interesting to deal with, because there’s also, presumably, like I said, I didn’t dig into it yet, but I will, like.
271 00:39:29.390 ⇒ 00:39:41.760 Katherine Bayless: exhibitors that are like, yes, we want this booth. No, actually, we want a bigger booth. No, actually, we want a smaller booth. Actually, also, we want this thing now. It’s like, I don’t know if, like, one exhibitor might have, like, 15 transactions from, you know.
272 00:39:41.760 ⇒ 00:39:54.099 Katherine Bayless: incrementally changing their footprint. And so, like, I’ll take a look at it, and I’ll see, too, if there are, like, questions that I can’t figure out. I can always ask Tom to kind of explain, like, how the heck does this work? So…
273 00:39:54.270 ⇒ 00:39:55.060 Katherine Bayless: Yeah.
274 00:39:56.060 ⇒ 00:39:56.880 Amber Lin: Cool.
275 00:39:57.430 ⇒ 00:40:14.759 Amber Lin: I had a few more things on my agenda, I’ll run through them real quick. So first is… I sent this stock in our channel. These are some open questions that I still have. I don’t think we have time to go over them in…
276 00:40:15.220 ⇒ 00:40:19.789 Amber Lin: this meeting. But I think the main one is…
277 00:40:20.500 ⇒ 00:40:25.820 Amber Lin: The on-site classification, the numbers just… Look, it’s…
278 00:40:26.150 ⇒ 00:40:30.869 Amber Lin: I’m just not 100% sure of the definitions.
279 00:40:30.870 ⇒ 00:40:31.520 Katherine Bayless: Yeah.
280 00:40:31.520 ⇒ 00:40:47.999 Amber Lin: hey, do we… because right now, what I’m… sorry, I’m not going to go in too deep, but we can go into it, if we… if we have time. So, these are just some questions. I shared this stock, I’ll share it in our chat again. I would love to get some…
281 00:40:48.190 ⇒ 00:40:49.410 Amber Lin: Comments?
282 00:40:49.510 ⇒ 00:40:51.549 Amber Lin: And then we feel…
283 00:40:51.550 ⇒ 00:41:07.320 Katherine Bayless: for the on-site one, I… I also wasn’t really able to get an exact answer. I have a feeling that… so there’s… there is in the data, I forget where it is, which column, but, like, somewhere in there, there is essentially a flag that’s, like.
284 00:41:07.410 ⇒ 00:41:25.560 Katherine Bayless: where did they register? And so, like, you would think that if the where did they register thing is on-site… yeah, there you go, yeah. You would think that, if that value is in there, it means they’re on-site, but it does not. It really, I think, is that, yeah, like, registration week equals zero thing?
285 00:41:25.560 ⇒ 00:41:39.500 Katherine Bayless: But I remember when I was working on this audit report, and I was like, it doesn’t matter what definition I try, I cannot reverse engineer the older numbers, and so the on-site thing, I think, is a little slippery, but yeah.
286 00:41:39.690 ⇒ 00:41:58.750 Amber Lin: Yeah, I… Quick overview. I think one is the flag, and one is the date range. Is on-site is derived… well, at least the robot tells me, is derived from registration interface, and it is from delegate. So, I don’t know why it’s called delegate, or what it might mean.
287 00:41:58.750 ⇒ 00:42:07.589 Amber Lin: So if we should include or exclude delegates that didn’t register on-site, but registered before, so…
288 00:42:08.100 ⇒ 00:42:12.250 Amber Lin: More confusion is there, and then it’s the…
289 00:42:12.530 ⇒ 00:42:22.750 Amber Lin: So this is including and excluding delegates, I’m getting slightly closer. And then there’s date range, because
290 00:42:22.950 ⇒ 00:42:30.619 Amber Lin: just registration week equals zero does not get me close enough to the numbers, but if I were to say,
291 00:42:31.400 ⇒ 00:42:42.369 Amber Lin: because media comes a bit earlier, so they might register on-site earlier. I don’t know if registration week equals zero is always accurate.
292 00:42:44.070 ⇒ 00:42:53.880 Katherine Bayless: Right. I don’t know that it is, and I have a funny feeling, like, I wouldn’t be surprised if, like, the date where, like, registration week equals zero is, like, some sort of strange thing that, like, counted backwards from, like.
293 00:42:54.050 ⇒ 00:42:58.690 Katherine Bayless: the Friday at the end of CES or something, like, I… yeah. But yeah, I had the same problem.
294 00:42:58.690 ⇒ 00:43:18.510 Amber Lin: it could be. So I tried different date ranges. Seems like for media, when I gave media an even longer time range, the numbers are a little bit closer, and I gave these, like, 7 days prior, and then the number is also somewhat closer, so especially when we have different
295 00:43:19.520 ⇒ 00:43:37.130 Amber Lin: timeframes of on-site for different registration types, it gets a little bit wonky of what day do we give it, what date do we give different registration types, so, like, it was just very hard to get, right? Anyways, the other ones are, like, less.
296 00:43:37.130 ⇒ 00:43:56.970 Katherine Bayless: probably should be included. They are a program where it can be technically a company or a country, but the idea is, like, one person’s gonna bring a whole crew. It’s like the group pass to CES, but you have to go through a whole bunch of very specific gates. And so the delegate registers for CES, and then they
297 00:43:57.770 ⇒ 00:44:10.230 Katherine Bayless: I actually really don’t even know how this works, to be totally honest. Somehow or other, they figure out how to, like… there’s, like, a delegate leader dashboard, and they can provision that to their people and tell them to go sign up there, and it comes through. I see.
298 00:44:11.060 ⇒ 00:44:11.900 Amber Lin: Yeah.
299 00:44:12.170 ⇒ 00:44:23.130 Amber Lin: Okay, we can… we can investigate more of the others, or more, like, definitions of, hey, what does this mean? What… what is the source of truth? This is what I can find.
300 00:44:23.130 ⇒ 00:44:37.549 Amber Lin: But I have the results, expected numbers, results, and the most… I try to include the queries I use, so if you guys have time, would love some comments here to close it out, and I think that would be…
301 00:44:37.660 ⇒ 00:44:45.230 Amber Lin: all of the pre-audit questions after we go through that. And if you want to see my very messy notes, it is down here.
302 00:44:45.440 ⇒ 00:44:47.640 Amber Lin: Do I recommend staying in here?
303 00:44:49.760 ⇒ 00:44:50.360 Katherine Bayless: So that?
304 00:44:51.790 ⇒ 00:44:53.060 Katherine Bayless: Did we ever get you a box?
305 00:44:54.220 ⇒ 00:44:54.940 Amber Lin: What the fuck?
306 00:44:54.940 ⇒ 00:44:55.490 Uttam Kumaran: Hmm.
307 00:44:55.730 ⇒ 00:45:00.190 Katherine Bayless: Oh, Access2Box, sorry, so that we could share some of the, like, additional reports and files.
308 00:45:00.190 ⇒ 00:45:05.910 Amber Lin: Oh, yeah, I would love that, because I want to see how the numbers match up.
309 00:45:06.070 ⇒ 00:45:10.839 Katherine Bayless: Yeah, I know, it’s funny, I don’t know where that got lost in translation.
310 00:45:11.980 ⇒ 00:45:17.770 Amber Lin: I think we stopped on being excited if we didn’t use, SharePoint.
311 00:45:17.770 ⇒ 00:45:21.699 Katherine Bayless: Right. Well, no, I think I… yeah, let me ask Ian.
312 00:45:21.700 ⇒ 00:45:22.350 Amber Lin: Okay.
313 00:45:25.050 ⇒ 00:45:36.589 Amber Lin: And in terms of rollout, I know Utam’s closing the sandbox. I think on my side, probably will work with…
314 00:45:36.730 ⇒ 00:45:48.910 Amber Lin: Jay or Ian on the roll chat, now that we have the Cortex agent. I’m working on the user guide, I’ll probably assign out some of the other…
315 00:45:49.010 ⇒ 00:45:55.359 Amber Lin: items, that’s different than the agent or the semantic views to folks on the team?
316 00:45:55.780 ⇒ 00:45:57.560 Awaish Kumar: Do we still need rule chat?
317 00:45:58.720 ⇒ 00:46:01.759 Amber Lin: Oh, you’re right, because we’re not.
318 00:46:01.760 ⇒ 00:46:04.609 Awaish Kumar: With the role developer, you can basically do…
319 00:46:05.230 ⇒ 00:46:09.209 Awaish Kumar: have access to ProdMods, you can go to the Cortex agent and…
320 00:46:09.210 ⇒ 00:46:12.799 Amber Lin: Oh, no, I mean for… for other folks, not for me.
321 00:46:13.360 ⇒ 00:46:17.589 Awaish Kumar: Yeah, for the end users also, they need access to the…
322 00:46:17.820 ⇒ 00:46:19.030 Amber Lin: Oh…
323 00:46:19.030 ⇒ 00:46:19.970 Awaish Kumar: to go.
324 00:46:21.330 ⇒ 00:46:24.110 Uttam Kumaran: But they just need to read access to Prod Martz.
325 00:46:25.290 ⇒ 00:46:26.030 Awaish Kumar: Yeah.
326 00:46:28.040 ⇒ 00:46:31.420 Uttam Kumaran: But role developer has, like, much more elevated permissions, right?
327 00:46:31.950 ⇒ 00:46:35.740 Awaish Kumar: Okay, so, like, I’m saying role chat was more, like, for the sidebar.
328 00:46:35.930 ⇒ 00:46:39.519 Awaish Kumar: Are we using that to give access to agent as well?
329 00:46:41.380 ⇒ 00:46:43.550 Amber Lin: I hope so, yeah, okay.
330 00:46:44.190 ⇒ 00:46:48.790 Amber Lin: I… I think we need to finalize it there.
331 00:46:48.960 ⇒ 00:46:59.259 Katherine Bayless: I mean, unless there’s any reason that we wouldn’t want to do that, but that would… my expectation was, yeah, roll chat would cover sidebar or agent, and any chat. Chat of any kind.
332 00:47:00.590 ⇒ 00:47:01.760 Amber Lin: And…
333 00:47:02.030 ⇒ 00:47:07.969 Awaish Kumar: I was thinking, like, we are more kind of restricting them to just being the agent, and not…
334 00:47:08.100 ⇒ 00:47:14.359 Awaish Kumar: Chat and on the side, because their numbers won’t matter if they chat in the agent, and they chat on the side.
335 00:47:14.820 ⇒ 00:47:25.900 Katherine Bayless: I see what you’re saying, I see what you’re saying. Should there be a… an even more restricted role that has access to only the agents, not even role chat in the SnowSite UI?
336 00:47:26.660 ⇒ 00:47:31.990 Amber Lin: We could maybe have the Snow Intelligence link in Okta, so they just click.
337 00:47:31.990 ⇒ 00:47:32.660 Katherine Bayless: Yeah.
338 00:47:32.660 ⇒ 00:47:35.239 Amber Lin: that, instead of going to Snowflake?
339 00:47:36.020 ⇒ 00:47:40.639 Katherine Bayless: That would be… because that’s a good… it’s a good point, Awish. I see now what you’re saying. I…
340 00:47:40.780 ⇒ 00:47:54.000 Katherine Bayless: I would still like to just use rolled chat, but I think, yeah, maybe the compromise is if we can just link them to Snowflake Intelligence to start, and then, yeah, like, if they hop over to the other side, that’s fine too, but…
341 00:47:54.280 ⇒ 00:48:01.929 Awaish Kumar: But, like, what I’m saying, we can have role chat, but they can see the database, and then they can just access the agent.
342 00:48:02.260 ⇒ 00:48:15.899 Awaish Kumar: the, like, should not be able to carry in the sidebar, because that’s the easy thing, right? Once you are in the table view, you see the sidebar, you just start asking there, instead of going to the… another tab for the agent.
343 00:48:16.240 ⇒ 00:48:19.580 Awaish Kumar: And then the numbers will, like, will break.
344 00:48:19.580 ⇒ 00:48:24.049 Uttam Kumaran: Yeah, I think I wish, I’d be interested to see if, like, if… can you only restrict access to just the agent?
345 00:48:24.290 ⇒ 00:48:28.330 Uttam Kumaran: Because, yeah, like, that’s the preferred surface versus using the sidebar.
346 00:48:28.530 ⇒ 00:48:29.280 Uttam Kumaran: Now.
347 00:48:30.150 ⇒ 00:48:31.000 Uttam Kumaran: Right.
348 00:48:31.130 ⇒ 00:48:31.880 Uttam Kumaran: I know.
349 00:48:32.070 ⇒ 00:48:40.749 Katherine Bayless: Yes, but we’d have to hurry up and get the… like, because the nice thing about the sidebar is there’s more data that’s in there, because.
350 00:48:40.750 ⇒ 00:48:41.660 Uttam Kumaran: Okay.
351 00:48:41.660 ⇒ 00:48:49.159 Katherine Bayless: I mean, I think… and then also there’s the streamlit piece, which… Yeah. Yeah.
352 00:48:51.390 ⇒ 00:49:00.510 Katherine Bayless: I mean, the sidebar chat should eventually be the same as the agents, I would think. I mean, the numbers shouldn’t be dramatically different.
353 00:49:01.130 ⇒ 00:49:05.630 Amber Lin: The numbers wouldn’t be that different, it’s mostly the routing between.
354 00:49:05.630 ⇒ 00:49:06.170 Katherine Bayless: topics.
355 00:49:06.170 ⇒ 00:49:13.389 Amber Lin: that might get caught. I think it will be much further in the future of all the definitions are
356 00:49:13.780 ⇒ 00:49:21.210 Amber Lin: actually in the tables themselves, that we can stay away from semantic views, but I don’t see a reason
357 00:49:21.490 ⇒ 00:49:27.509 Amber Lin: Why not use the agent? Because it has better, wrapping, it has just…
358 00:49:27.920 ⇒ 00:49:30.520 Amber Lin: Formatting and routing and all of that.
359 00:49:31.320 ⇒ 00:49:37.830 Awaish Kumar: Yeah, so, like, the way Snowflake… like, why Snowflake has the concept of semantic views is…
360 00:49:38.050 ⇒ 00:49:43.549 Awaish Kumar: It’s because of this… they want to control the… How… how are you…
361 00:49:43.790 ⇒ 00:49:46.890 Awaish Kumar: You interact with the data, and how it’s supposed to give the answer.
362 00:49:47.370 ⇒ 00:49:48.269 Awaish Kumar: Yeah, yeah.
363 00:49:48.420 ⇒ 00:49:51.649 Katherine Bayless: I would say, okay, I guess what I… to say it differently.
364 00:49:51.940 ⇒ 00:50:14.600 Katherine Bayless: I would be fine with making roll chat streamlit and agents only, but only if we hurry up and get agents for all of the things they can… because I can’t take away things that they have, right? Like, that’s the… that’s kind of the challenge, is, like, there’s a lot of people who are already using the sidebar chat, where they can get to the membership data and some of the other stuff that we’ve, you know, modeled as far as the marts, but we don’t have the semantic view on top of.
365 00:50:14.600 ⇒ 00:50:23.130 Katherine Bayless: So, like, until we have semantic views so that they can have agents for all those other data sources, I can’t… I can’t take away the fun they’re having.
366 00:50:23.130 ⇒ 00:50:30.309 Amber Lin: Quick note here, I did add an agent. You can use it in an agent site called Last Restore. It does have.
367 00:50:30.310 ⇒ 00:50:30.740 Katherine Bayless: access.
368 00:50:30.740 ⇒ 00:50:45.259 Amber Lin: two semantic views and the ProdMarch data, so they can query it in there. It just… the agent will say, hey, you only use this, as last resort, the definition is not finalized, but it will be able to query the data directly.
369 00:50:45.760 ⇒ 00:50:52.960 Katherine Bayless: Okay, okay. If that’s possible, then I think maybe we’d call it something other than last resort all the way.
370 00:50:53.640 ⇒ 00:51:07.980 Katherine Bayless: get why we’re calling it that on our side. But yeah, so that’s kind of the generic agent, sort of, that I had asked about, I guess. Okay, okay, okay. See, I think, I mean, yeah, as long as they can get to the same information.
371 00:51:08.000 ⇒ 00:51:15.819 Katherine Bayless: then I don’t mind cutting off the sidebar. I just wouldn’t want anybody to be like, you know, dang it, I was working on that stuff.
372 00:51:15.900 ⇒ 00:51:19.509 Katherine Bayless: Yeah, that’s all. But yeah, if there’s the generic agent, then I think that’s fine.
373 00:51:22.940 ⇒ 00:51:23.580 Awaish Kumar: Okay.
374 00:51:24.190 ⇒ 00:51:28.760 Katherine Bayless: We will have to tell them to go to the new place, but we… Yeah. Yeah, yeah.
375 00:51:29.520 ⇒ 00:51:38.510 Katherine Bayless: We can do that, that’s not a problem. And I do like the idea of changing the, like, default Octolink to being the Snowflake Intelligence link.
376 00:51:38.510 ⇒ 00:51:42.530 Uttam Kumaran: Yeah, I kind of like that, too. I don’t know, also, we can ask Ian…
377 00:51:42.770 ⇒ 00:51:48.250 Uttam Kumaran: Amber, like, if we can draw people into there, Initially, I don’t know.
378 00:51:49.270 ⇒ 00:51:55.860 Uttam Kumaran: is, like, not something I… I immediately asked about on our first call, but now that I’m thinking about it, yeah, I don’t know if we can do the redirect.
379 00:51:56.300 ⇒ 00:51:58.420 Uttam Kumaran: That would be actually so convenient.
380 00:51:59.190 ⇒ 00:52:07.689 Katherine Bayless: I know, I think… I do think we can… actually, I think I have some of the ability to configure this Trellica thing.
381 00:52:09.990 ⇒ 00:52:26.170 Katherine Bayless: Yeah, yeah, yeah, so we can do an application URL and a sign-on URL, and, like, they can be different. So as long as there is a, like, deep link for the Snowflake Intelligence stuff that will flow through the auth, then yeah, it should be totally doable.
382 00:52:29.690 ⇒ 00:52:33.889 Katherine Bayless: That’d be cool. Then the only problem becomes, how do we get them from there to Streamlit, but…
383 00:52:34.460 ⇒ 00:52:35.640 Katherine Bayless: Feels solvable.
384 00:52:36.770 ⇒ 00:52:44.639 Amber Lin: maybe a different app, a different link for Streamlight Apps, because I feel like there are different users who’s using it.
385 00:52:45.000 ⇒ 00:52:58.900 Amber Lin: And also in, in Snowflake Intelligence, there’s artifacts where you can save previous queries, so if they only want one thing and not the whole dashboard, they can just save that query
386 00:52:59.090 ⇒ 00:53:04.149 Amber Lin: In artifacts, and go look at exhibitor countries, or whatever.
387 00:53:04.370 ⇒ 00:53:05.200 Katherine Bayless: Yeah.
388 00:53:05.470 ⇒ 00:53:09.659 Katherine Bayless: Yeah, actually, I like the idea of the two tiles,
389 00:53:10.060 ⇒ 00:53:21.940 Katherine Bayless: That, interestingly, is when I’m like, I don’t know if that’s possible, but I don’t see why not. I mean, I think in… as far as Okta’s concerned, you’re just defining a tile, and a tile is just a, like, URL, so… yeah.
390 00:53:22.910 ⇒ 00:53:27.980 Katherine Bayless: Okay, I like that idea. I think that’d be worth… yeah, and then we could have Streamlit and…
391 00:53:28.530 ⇒ 00:53:30.820 Katherine Bayless: Robot, data robot.
392 00:53:32.750 ⇒ 00:53:33.899 Katherine Bayless: I like that.
393 00:53:35.300 ⇒ 00:53:40.080 Amber Lin: Also, me and Utam have booked our flights, so we’ll be there…
394 00:53:40.380 ⇒ 00:54:00.300 Amber Lin: 12th and the 13th? I think I’ll be there full day on the 12th. I land, like, 5 or 6 in the morning, and then I’m leaving 13… my flight out is the 13th at 5, so I have pretty much a day and a half fully. And then, Utam, I…
395 00:54:00.300 ⇒ 00:54:04.730 Uttam Kumaran: Something similar, I’m there, like, sometime in the morning, and then I’m leaving, like, 6-something.
396 00:54:04.860 ⇒ 00:54:05.840 Uttam Kumaran: Wednesday.
397 00:54:06.160 ⇒ 00:54:08.170 Katherine Bayless: Nice, nice, okay.
398 00:54:08.170 ⇒ 00:54:11.499 Amber Lin: So we can plan those sessions or book the rooms if we need.
399 00:54:12.180 ⇒ 00:54:17.919 Katherine Bayless: Okay, perfect. Do we… do we wanna do that now? Or do we wanna…
400 00:54:17.920 ⇒ 00:54:18.650 Uttam Kumaran: Yeah.
401 00:54:18.650 ⇒ 00:54:19.190 Amber Lin: Sure.
402 00:54:19.980 ⇒ 00:54:32.119 Katherine Bayless: Okay, so… in terms of… because I think, if I’m remembering correctly, we wanted to have, like, an office hours-y kind of thing, lunch and learn-y kind of thing, so we could do…
403 00:54:32.280 ⇒ 00:54:36.260 Katherine Bayless: Wednesday, we could do that, like,
404 00:54:36.670 ⇒ 00:55:01.610 Katherine Bayless: after the division meeting ends, which I think, if you guys wanted to drop by, that too, because I am… I’m pretty sure I signed up to do a snowflake demo at that thing. So yeah, if you wanted to do the division meeting, that starts at 9.30 and goes for 2 hours, so we can… I can figure out where in there that the snowflakes piece will happen, and I could ask for it to be closer to the end, and then we could kind of, like, go from that into Lunch and Learn slash office hours, if the
405 00:55:01.610 ⇒ 00:55:07.040 Katherine Bayless: Like, 11 to maybe 2 o’clock block is okay? On the 13th.
406 00:55:09.310 ⇒ 00:55:10.310 Uttam Kumaran: That’s perfect.
407 00:55:10.550 ⇒ 00:55:11.090 Katherine Bayless: Yep.
408 00:55:11.260 ⇒ 00:55:12.100 Uttam Kumaran: Yeah, let’s do that.
409 00:55:12.350 ⇒ 00:55:12.910 Katherine Bayless: Okay.
410 00:55:13.030 ⇒ 00:55:20.379 Katherine Bayless: And then, like, on the 12th, we can hang out, but we can make, like, the sort of broader, like, group thing on the… yeah, okay.
411 00:55:20.600 ⇒ 00:55:21.140 Uttam Kumaran: Okay.
412 00:55:24.010 ⇒ 00:55:35.090 Katherine Bayless: So if you wanted to come to the Snowflake, or the, yeah, the Snowflake demo to CES folks, as flies on walls, that is the 12th, 12th, 12th to 12th.
413 00:55:35.320 ⇒ 00:55:38.960 Katherine Bayless: So yeah, that one’s up to you guys, I mean, honestly, like…
414 00:55:39.980 ⇒ 00:55:52.850 Katherine Bayless: it’s a lot of things, not just the Snowflake demo, and I think part of the reason that I was a little late to be on the call and paying attention today is, likely to wind up dominating that conversation anyway, because it’s really kind of…
415 00:55:52.970 ⇒ 00:55:54.949 Katherine Bayless: That group is very…
416 00:55:55.610 ⇒ 00:56:04.040 Katherine Bayless: they’re probably gonna wind up being some of our tip-of-the-spear people on these AI sort of projects, and right now, I think everybody’s trying to figure out, like.
417 00:56:04.530 ⇒ 00:56:14.820 Katherine Bayless: you know, not that people don’t love me and Jay, but I think everybody’s like, okay, well, who’s gonna be that guinea pig? Yeah. Like, I don’t know, we’re in some interesting conversations in the moment.
418 00:56:14.820 ⇒ 00:56:27.959 Katherine Bayless: So, like, I… the CES meeting on the Tuesday might wind up being more about the, like, AI project kind of stuff than the Snowflake stuff, even if we have the demo, but, still up to you guys if you want to just kind of, like, come and listen in.
419 00:56:28.530 ⇒ 00:56:32.409 Uttam Kumaran: I would… I would love to listen in with you, Amber, that’d be great.
420 00:56:33.230 ⇒ 00:56:37.899 Katherine Bayless: Okay, okay. I will… I’ll clear it with JTK, just to be polite, but it shouldn’t be a problem.
421 00:56:37.900 ⇒ 00:56:38.450 Uttam Kumaran: Okay.
422 00:56:39.940 ⇒ 00:56:42.350 Katherine Bayless: And then I’ll forward you the invite if he’s good with it.
423 00:56:45.970 ⇒ 00:56:46.660 Katherine Bayless: Okay.
424 00:56:48.840 ⇒ 00:57:13.650 Katherine Bayless: Well, I know we’re more or less at the hour, so everybody’s gotta go, but, the only other thing that was just kind of generally on my mind was, going back to the streamlit thing. It’s like, I think… I know Kai’s kind of started down this path of, like, getting the apps that we sort of had built prior to having the fancy way of doing it, kind of cleaned up, but I… I continue to keep creating, tech debt, with these things, and so I think a little streamlit clean
425 00:57:13.650 ⇒ 00:57:16.110 Katherine Bayless: up to before we get into some of the, like.
426 00:57:16.570 ⇒ 00:57:33.939 Katherine Bayless: rollout stuff, would be good. Just making sure, especially if we are gonna drive people more so to the intelligence side, like, like, making sure those are connected in well. Like, do we need to co-locate the Streamlit apps with the semantic views that they’re relevant to, or things like that?
427 00:57:34.540 ⇒ 00:57:35.100 Uttam Kumaran: Okay.
428 00:57:35.430 ⇒ 00:57:45.359 Amber Lin: I can also look into if we can add that as a tool, perhaps, or at least as links in responses, so people can click on the link and…
429 00:57:45.890 ⇒ 00:57:49.610 Amber Lin: transport over, I can look into that.
430 00:57:49.750 ⇒ 00:57:52.980 Awaish Kumar: Yeah, like… Table of content.
431 00:57:53.390 ⇒ 00:57:58.930 Awaish Kumar: Simulator app, which will list all the apps, and then you can go from there.
432 00:58:00.800 ⇒ 00:58:24.670 Katherine Bayless: That would be cool, too, because potentially it would give us a chance… because the other thing that Kai and I have talked about is, like, well, the information architecture on this stuff is going to get crazy fast, like, if we’ve got all these coming in, and right now, at least in the UI, there’s not really much of a, like, file structure kind of vibe, they’re just a long list with some filters, which is fine, but if we had this sort of set up, like, that way, Oish, we could probably
433 00:58:24.740 ⇒ 00:58:29.610 Katherine Bayless: Inject a little bit of information architecture, in the middle, which would help, yeah.
434 00:58:29.740 ⇒ 00:58:30.600 Katherine Bayless: Yeah.
435 00:58:30.900 ⇒ 00:58:32.420 Katherine Bayless: Okay. I like that.
436 00:58:32.420 ⇒ 00:58:35.750 Awaish Kumar: We can explore that, if that’s possible, yeah.
437 00:58:36.420 ⇒ 00:58:37.030 Katherine Bayless: Okay.
438 00:58:37.690 ⇒ 00:58:38.480 Katherine Bayless: Cool.
439 00:58:39.120 ⇒ 00:58:50.229 Katherine Bayless: And then informational update, the AWS stuff, I think, Ian, deployed his first account yesterday, and so I think, we’ll be able to start…
440 00:58:50.580 ⇒ 00:59:09.860 Katherine Bayless: Migrating all of this stuff. Next week, we are still waiting to hear on the Snowflake detach, reattach, piece for the production one, but like I said, I’m not super worried. Worst case scenario, we just run all the code in a new Snowflake instance. But ideally, we’ll be able to detach and reattach it to the new accounts.
441 00:59:11.060 ⇒ 00:59:13.979 Awaish Kumar: Okay, and what’s the status on Stripe thing?
442 00:59:14.370 ⇒ 00:59:35.229 Katherine Bayless: Mmm, stripe, it’s a good question. So we met with, the sales team yesterday? Yeah, yesterday. And actually, it went really well. They were really excited, which, is… it was cool. It was a… it was a cool moment. It was a good counterbalance to the, like, afternoon meeting that was a little bit more hostile with a different team. But, they’re on board. What we’re gonna do with it is…
443 00:59:35.230 ⇒ 00:59:43.460 Katherine Bayless: use it to sell the sponsorships for the LIT dinner, so the sales will come online, probably.
444 00:59:43.470 ⇒ 00:59:47.840 Katherine Bayless: June-ish, and in the meantime, we’re gonna work with them, well…
445 00:59:47.860 ⇒ 01:00:07.060 Katherine Bayless: Royal we, I don’t have a frog in my pocket. Jackie is gonna work with them on figuring out how to use the Salesforce product catalog, tie that to Stripe, and that way we don’t have to maintain a product catalog for them, they can just define it in Salesforce, right? Same place they’re already working, and then they can grab the URLs and send them to people for payment.
446 01:00:07.060 ⇒ 01:00:16.609 Katherine Bayless: So they’re gonna start scoping all of that out. So I don’t think we’ll have data coming in through Stripe for a while, but at the very least, we are advancing the, the, idea, which is cool.
447 01:00:17.620 ⇒ 01:00:25.630 Katherine Bayless: So yeah, cautiously optimistic that we’ll be able to, like, do well with the little pilot, and then everybody will be like, this is great, use it for all the things! So…
448 01:00:26.380 ⇒ 01:00:27.070 Katherine Bayless: Yeah.
449 01:00:27.410 ⇒ 01:00:27.920 Uttam Kumaran: Damn.
450 01:00:28.270 ⇒ 01:00:29.680 Katherine Bayless: Yeah, we’ll see what happens.
451 01:00:34.390 ⇒ 01:00:39.250 Uttam Kumaran: And then I’m gonna go ahead and execute the sandbox breakdown today.
452 01:00:39.790 ⇒ 01:00:40.859 Katherine Bayless: Yeah, I…
453 01:00:40.860 ⇒ 01:00:43.149 Uttam Kumaran: I looked through it, there wasn’t anything there, so…
454 01:00:43.520 ⇒ 01:00:46.600 Katherine Bayless: I think the problem is… so yeah, there’s nothing in there, totally.
455 01:00:46.600 ⇒ 01:00:46.920 Uttam Kumaran: Okay.
456 01:00:47.330 ⇒ 01:00:52.510 Katherine Bayless: I think the problem is I tried to cut it… shut it down too, but I think we might have to, like, reach out to their support to do it?
457 01:00:52.510 ⇒ 01:00:53.909 Uttam Kumaran: Yeah, I will, I will.
458 01:00:53.910 ⇒ 01:00:55.920 Katherine Bayless: Okay, okay, okay, okay. That was right.
459 01:00:55.920 ⇒ 01:00:57.780 Uttam Kumaran: There’s no turn-off button, yeah, I have to…
460 01:00:57.780 ⇒ 01:01:00.490 Katherine Bayless: Like, I just wanted the delete button. I guess it makes sense that there.
461 01:01:00.490 ⇒ 01:01:08.210 Uttam Kumaran: Yeah, they’re pretty quick, but I just… I’m gonna… I just looked through it today, and ran a little script and just made sure that there’s nothing there, so…
462 01:01:08.210 ⇒ 01:01:13.310 Katherine Bayless: Okay, okay, yeah, that was where I got stuck, was the, I was like, oh god, I gotta reach out to somebody?
463 01:01:14.060 ⇒ 01:01:15.980 Katherine Bayless: Terrible.
464 01:01:16.540 ⇒ 01:01:17.270 Katherine Bayless: Terrible.
465 01:01:17.790 ⇒ 01:01:19.840 Katherine Bayless: But yeah.
466 01:01:22.040 ⇒ 01:01:22.700 Uttam Kumaran: Okay.
467 01:01:23.240 ⇒ 01:01:23.840 Awaish Kumar: Okay.
468 01:01:25.000 ⇒ 01:01:26.590 Uttam Kumaran: Thank you, everyone. Appreciate it.
469 01:01:27.400 ⇒ 01:01:28.080 Katherine Bayless: Talk to you soon.
470 01:01:28.290 ⇒ 01:01:28.760 Awaish Kumar: Enjoy.
471 01:01:28.760 ⇒ 01:01:29.469 Katherine Bayless: Every weekend?
472 01:01:29.470 ⇒ 01:01:31.319 Amber Lin: See you soon. Bye. Bye.