2026-03-26_cta_x_brainforge_weekly_sync

Meeting Title: CTA x Brainforge Weekly Sync Date: 2026-03-26 Meeting participants: Awaish Kumar, Uttam Kumaran, Katherine Bayless, Kyle Wandel

WEBVTT

1 00:00:05.850 ⇒ 00:00:06.590 Awaish Kumar: Hi.

2 00:00:12.980 ⇒ 00:00:18.489 Uttam Kumaran: Hey, Awash, I’m just gonna, run to the restroom, just been, like, back-to-back, so give me one sec.

3 00:01:23.740 ⇒ 00:01:26.010 Katherine Bayless: Hello, hello, hello. Give me, like, 2 minutes.

4 00:01:26.570 ⇒ 00:01:31.419 Katherine Bayless: Just been on back-to-back calls, so just gonna make a quick restroom break. I’ll be right back.

5 00:02:22.280 ⇒ 00:02:22.940 Awaish Kumar: Okay.

6 00:02:24.500 ⇒ 00:02:25.490 Uttam Kumaran: Hey, sorry.

7 00:03:22.800 ⇒ 00:03:25.250 Katherine Bayless: Okay, sorry guys, thank you.

8 00:03:25.250 ⇒ 00:03:26.220 Uttam Kumaran: Oh, God.

9 00:03:27.150 ⇒ 00:03:28.760 Katherine Bayless: Back-to-back calls all day.

10 00:03:28.760 ⇒ 00:03:37.770 Uttam Kumaran: Me too, I was just, I was just like, let me… I just, I hopped on, I was like, oh wait, I need to just, like, go literally grab, like, snack from the fridge, go to the bathroom.

11 00:03:38.120 ⇒ 00:03:44.550 Katherine Bayless: Exactly, exactly. The only difference is my fridge has no snacks, which will be noted in its performance review.

12 00:03:44.550 ⇒ 00:03:47.100 Uttam Kumaran: I also didn’t find any stat.

13 00:03:47.760 ⇒ 00:03:52.490 Uttam Kumaran: I just… I didn’t find any snacks, I was like… I’m, like, digging through there, like…

14 00:03:53.510 ⇒ 00:03:56.909 Uttam Kumaran: Who, like, who works around here?

15 00:03:56.910 ⇒ 00:03:57.750 Katherine Bayless: Who knows?

16 00:03:57.750 ⇒ 00:03:59.510 Uttam Kumaran: Run in this operation.

17 00:04:00.090 ⇒ 00:04:03.600 Katherine Bayless: Yeah, exactly. I have no one to blame but myself, but I’ll find someone.

18 00:04:03.600 ⇒ 00:04:13.669 Uttam Kumaran: Yeah, it’s like, it’s like, have you seen the… the memes where, like, you, like, look in the mirror, and it’s, like, it’s just, like, a clown, and you’re just looking at yourself? Yeah.

19 00:04:13.670 ⇒ 00:04:22.880 Katherine Bayless: exactly. But, of all the calls I’ve been stuck on today, this one I’m actually very excited about, so, you know.

20 00:04:22.880 ⇒ 00:04:30.180 Uttam Kumaran: Yeah, I have some very cool things, I don’t know, Waish, if you want to start with the ID stitching, I can…

21 00:04:30.390 ⇒ 00:04:33.299 Uttam Kumaran: I can pull my stuff up as well.

22 00:04:33.810 ⇒ 00:04:35.320 Katherine Bayless: No worries, wherever you guys want to start.

23 00:04:35.320 ⇒ 00:04:45.550 Awaish Kumar: Yeah, actually, I… we started with the… with the follow-ups on the CES modeling, and then the Fortune 500 and twice retail stuff.

24 00:04:45.940 ⇒ 00:04:58.810 Awaish Kumar: So now, all of these are added as a DIM table, dimension tables. We didn’t try to stitch it, with the registration data, because I remember, like, that we didn’t want it to happen.

25 00:04:59.010 ⇒ 00:05:12.629 Awaish Kumar: have some, like, the connection between them, for now at least. So we are keeping it separately as a dimension table, which can be utilized in future if needed, but for now, it’s just some company names and…

26 00:05:12.780 ⇒ 00:05:24.099 Awaish Kumar: And whatever we find in raw data. So… and apart from that, like, the exhibitor data has been modeled, so now this DIM exhibitor data is also coming from…

27 00:05:24.320 ⇒ 00:05:37.739 Awaish Kumar: that. So these are the few changes that we put in the same PR that I opened quite a few days back. I’m just waiting for someone to prove when we can merge it, and yeah, we are good on that.

28 00:05:38.160 ⇒ 00:05:49.149 Katherine Bayless: Okay, cool. Yeah, I think Tyle was working on kind of reviewing some of the specific details, but, because I just haven’t had the time to get into it too deeply, but it looked good at first glance.

29 00:05:50.920 ⇒ 00:06:04.409 Katherine Bayless: And I think actually next week, we’re gonna start the work on migrating the old, exhibitor sort of, like, pipeline that was used for reporting, and so we should start to be able to have live data by the end of next week.

30 00:06:04.550 ⇒ 00:06:11.370 Katherine Bayless: Coming through for that, because believe it or not, we are already selling space. We have been since we were on-site this year.

31 00:06:11.370 ⇒ 00:06:12.060 Uttam Kumaran: Wow.

32 00:06:12.450 ⇒ 00:06:13.030 Katherine Bayless: Yeah.

33 00:06:13.230 ⇒ 00:06:18.979 Katherine Bayless: Yeah, I forget how much they did tell us, like, how many millions of dollars we sold on-site at.

34 00:06:18.980 ⇒ 00:06:19.609 Uttam Kumaran: On the spot.

35 00:06:19.970 ⇒ 00:06:20.820 Katherine Bayless: Yeah.

36 00:06:21.190 ⇒ 00:06:21.990 Katherine Bayless: Yeah.

37 00:06:22.950 ⇒ 00:06:29.810 Awaish Kumar: The data, when you say live, it will be coming from… Some platform, or through the…

38 00:06:30.020 ⇒ 00:06:32.379 Awaish Kumar: The functions we have been writing, and…

39 00:06:32.630 ⇒ 00:06:51.400 Katherine Bayless: So it’ll be coming from, the platform we use is called ExpoCAD, and there’s an existing reporting pipeline that’s in the old AWS account that Jay and the IT team had written many years ago, and so we’re basically gonna rewrite the pipeline, right, give it a little love.

40 00:06:51.760 ⇒ 00:06:59.680 Katherine Bayless: And then it’ll be connecting directly to that platform and bringing the data into S3, where we can then take it from there into Snowflake.

41 00:06:59.820 ⇒ 00:07:06.029 Katherine Bayless: I mean, you know, just a fun adventure that’ll be on the table if anybody’s interested. We do also…

42 00:07:06.560 ⇒ 00:07:23.509 Katherine Bayless: We have also historically, I should say it that way, pushed this report that Jay’s Code would put together out via Slack, and I have a suspicion that we will have to continue to provide the report via Slack, even though it’ll also be in Snowflake, just because there’ll be people that aren’t

43 00:07:23.510 ⇒ 00:07:32.299 Katherine Bayless: in Snowflake yet that need the report, so, like I said, that part already is up and running and works, but we will get a chance to have our first, sort of, like.

44 00:07:32.480 ⇒ 00:07:35.349 Katherine Bayless: pipe out to Slack from our account and stack.

45 00:07:36.330 ⇒ 00:07:37.050 Awaish Kumar: Okay.

46 00:07:37.050 ⇒ 00:07:37.630 Uttam Kumaran: Great.

47 00:07:38.180 ⇒ 00:07:38.850 Katherine Bayless: Yeah.

48 00:07:41.320 ⇒ 00:07:42.680 Katherine Bayless: But yeah, that’s awesome.

49 00:07:44.040 ⇒ 00:07:44.700 Uttam Kumaran: Okay.

50 00:07:45.230 ⇒ 00:07:47.780 Uttam Kumaran: Oh, Wish, anything else on, like, ID stitching?

51 00:07:49.640 ⇒ 00:07:56.209 Awaish Kumar: Like, for the ID switching, right, we… I think we haven’t…

52 00:07:56.420 ⇒ 00:08:02.750 Awaish Kumar: like, the waterfall thing that we implemented, I think that’s what is in there.

53 00:08:03.680 ⇒ 00:08:08.889 Awaish Kumar: And we… yeah, I think we had some feedback from the last call, but yeah, there needs to be done.

54 00:08:09.730 ⇒ 00:08:13.219 Uttam Kumaran: Okay. Yeah, on the last call, I just… I asked if we can,

55 00:08:13.350 ⇒ 00:08:17.150 Uttam Kumaran: We talked about, like, also layering in Cortex for further.

56 00:08:17.410 ⇒ 00:08:19.869 Uttam Kumaran: So I think Ashwini is working on that, so…

57 00:08:19.990 ⇒ 00:08:26.100 Uttam Kumaran: Yeah, that’d be great to follow up on, so that we can, like, sort of put a bow in, I think, the first version of the logic.

58 00:08:27.490 ⇒ 00:08:37.880 Katherine Bayless: Did we… forgive me for not remembering off the top of my head, but did we go ahead and assign or otherwise devise a means of assigning the canonical IDs to the companies?

59 00:08:40.030 ⇒ 00:08:42.219 Uttam Kumaran: I don’t think so.

60 00:08:42.640 ⇒ 00:08:43.340 Katherine Bayless: Okay.

61 00:08:43.570 ⇒ 00:08:53.639 Katherine Bayless: we don’t… not to say that that’s, like, urgent due right now, but, the reason I’m thinking about it is, like, so with some of the stuff that we’re gonna be integrating for CES, we’re…

62 00:08:53.670 ⇒ 00:09:09.330 Katherine Bayless: my current expectation is that we will, because everybody has to authenticate through Okta, we will persist an ID for the person, and if applicable, their canonical company via Okta, and that way, those IDs can follow people kind of around the different systems.

63 00:09:09.330 ⇒ 00:09:14.930 Katherine Bayless: That we’re passing the data for CES stuff, and so if we have… and I know that I…

64 00:09:14.980 ⇒ 00:09:33.160 Katherine Bayless: the canonical IDs for people will be messy, because we’re not really doing the same level of entity resolution on them yet. So it’ll probably just be, like, one for each email address, essentially. But at least for the company ones, if we’ve got clean IDs, we can pair up to people. That makes sense to just kind of float those around, too. Then they’re there if we need them.

65 00:09:33.410 ⇒ 00:09:33.970 Uttam Kumaran: Okay.

66 00:09:35.460 ⇒ 00:09:53.770 Katherine Bayless: Oh, also because, right, sorry, the other piece of that puzzle is we are looking at, this year, a more push, like, more outreach for pre-credentialing folks. So, like, when somebody wants to go to CES, right, you have to, like, prove that you’re in the industry, which is ironic at this point in time. Who’s not in the tech industry, right?

67 00:09:53.770 ⇒ 00:09:54.270 Uttam Kumaran: Yes.

68 00:09:54.410 ⇒ 00:10:06.799 Katherine Bayless: And so, this year, we’ve always done it where, like, if you’re a member company, we, you know, we know you’re in the industry, and so we will pre-approve your credentials for CES, but we’re gonna open that up, to, like.

69 00:10:07.390 ⇒ 00:10:16.880 Katherine Bayless: Other companies basically being able to come in, fill out a form, submit their information, and then, if approved, anybody that comes in with, like, their email domain, we would also have.

70 00:10:16.880 ⇒ 00:10:17.420 Uttam Kumaran: Oh, okay.

71 00:10:17.420 ⇒ 00:10:28.339 Katherine Bayless: pre-credentialed for CES, and so the canonical company IDs will also matter there. But, like, it’s not urgent. I mean, I think that work will start in the next, you know, 6 weeks or so, so we got a little time.

72 00:10:29.160 ⇒ 00:10:29.770 Katherine Bayless: Yeah.

73 00:10:29.770 ⇒ 00:10:30.310 Uttam Kumaran: Okay.

74 00:10:32.660 ⇒ 00:10:33.749 Katherine Bayless: And we can pitch it. Cool.

75 00:10:33.980 ⇒ 00:10:35.359 Katherine Bayless: Yeah, yeah, yeah, yeah, yeah.

76 00:10:35.360 ⇒ 00:10:43.220 Uttam Kumaran: Okay, yeah, that’d be great. I think maybe, Awash, that would be something that I would like… we just try to close out this week, is the remaining on the ID stitching.

77 00:10:43.660 ⇒ 00:10:48.629 Uttam Kumaran: And then I made some interesting progress on…

78 00:10:48.800 ⇒ 00:10:51.889 Uttam Kumaran: Cortex adoption, so I can share…

79 00:10:52.020 ⇒ 00:10:58.410 Uttam Kumaran: what I have so far, and as I’ll share the output, and I’ll back into, like, how I got there, and then hopefully…

80 00:10:58.820 ⇒ 00:11:04.079 Uttam Kumaran: I can… Probably help everybody here set up, like, sort of the dev process.

81 00:11:04.270 ⇒ 00:11:09.329 Uttam Kumaran: So… Yes, I was… so…

82 00:11:09.960 ⇒ 00:11:16.080 Uttam Kumaran: There are multiple Cortex products. There’s Cortex Agent, there’s Cortex Analyst, and there’s Cortex CLI.

83 00:11:16.250 ⇒ 00:11:21.140 Uttam Kumaran: It was looking at… Vortex CLI usage.

84 00:11:21.430 ⇒ 00:11:22.140 Katherine Bayless: Which…

85 00:11:22.140 ⇒ 00:11:26.580 Uttam Kumaran: It’s not named. None of the names of the tables in the information schema indicate

86 00:11:26.770 ⇒ 00:11:37.120 Uttam Kumaran: the product, it’s just they’re… they’re, like, kind of off, so I was, like, digging through the docs, and then I… so I found all the usage. For now, I’ve really just focused on,

87 00:11:37.620 ⇒ 00:11:39.199 Uttam Kumaran: the sidebar. So, but…

88 00:11:39.920 ⇒ 00:11:46.209 Uttam Kumaran: In case we eventually want to measure… like, there’s the notion of agents, and there’s also the Cortex CLI, but…

89 00:11:46.380 ⇒ 00:12:00.560 Uttam Kumaran: for this piece, I was able to get all the data, and I basically, like, threw together a first draft of, like, a dashboard. I think we’re gonna have to definitely play around with, like, the viz and, like, some of the Python, but, like.

90 00:12:01.000 ⇒ 00:12:07.359 Uttam Kumaran: with vanilla, I feel like this is, like, a pretty good space to start. So…

91 00:12:07.910 ⇒ 00:12:11.440 Uttam Kumaran: So basically, like, I sort of showed, okay, like.

92 00:12:11.580 ⇒ 00:12:21.909 Uttam Kumaran: Are we seeing requests? Like, who’s actively using it? There’s some things that I need to just continue doing some investigations on. I’m like, okay, this doesn’t seem right when we have requests, but…

93 00:12:22.450 ⇒ 00:12:28.509 Uttam Kumaran: It’s been really tough to go from, like, modeling to then the Python for the app.

94 00:12:28.820 ⇒ 00:12:30.400 Uttam Kumaran: And, like, hook everything up.

95 00:12:30.640 ⇒ 00:12:36.319 Uttam Kumaran: So, I’m, like, just… that’s actually, like, the part I’m, like, trying to figure out broad, like, that’s the meta challenge here.

96 00:12:36.380 ⇒ 00:12:37.380 Katherine Bayless: Yeah.

97 00:12:37.600 ⇒ 00:12:48.009 Katherine Bayless: But, yeah. Same suffering. It’s funny, the member engagement report that we’ve got in Streamlit is, like, a little too big for Coco to, like, edit holistically, and so, yeah.

98 00:12:48.010 ⇒ 00:12:48.600 Uttam Kumaran: Yeah.

99 00:12:48.950 ⇒ 00:12:52.860 Katherine Bayless: editing it with Claude code externally, but then needs to, like.

100 00:12:52.860 ⇒ 00:13:04.839 Uttam Kumaran: And then copying and pasting it, and then there’s an error. So that’s what I think I’ve figured it out. I’ll show you the deploy script that I have that allows you to edit it in a local environment, and then send it up.

101 00:13:04.990 ⇒ 00:13:19.560 Uttam Kumaran: And then if there’s an issue, it’s able to actually, like, see that, and you can do some iteration. So then I have just, like, okay, who’s in the sidebar? I don’t know yet about, like, the access formatting. I think we will just need

102 00:13:19.880 ⇒ 00:13:26.140 Uttam Kumaran: Some other library for me to do some of, like, the intense… Viz cleanup, but, like.

103 00:13:26.250 ⇒ 00:13:32.290 Uttam Kumaran: Top 10 requests, who’s using it, tokens used, so all this data looks, like, fairly accurate.

104 00:13:32.580 ⇒ 00:13:40.019 Uttam Kumaran: You are winning. Wait, you’re winning? You’re winning in terms of requests, but… Kai’s questions may be beefier.

105 00:13:40.210 ⇒ 00:13:42.349 Uttam Kumaran: On average.

106 00:13:42.390 ⇒ 00:13:44.170 Katherine Bayless: Interesting. Interesting.

107 00:13:44.170 ⇒ 00:13:50.559 Kyle Wandel: Catherine, I don’t mean to, bother, but, I was late because we’re working on the… did you see that email from Michael Brown?

108 00:13:51.200 ⇒ 00:13:54.310 Katherine Bayless: No, sorry, I haven’t actually checked my email in a while.

109 00:13:54.310 ⇒ 00:14:11.290 Kyle Wandel: No, you’re good. We wanted updated numbers for twice and, buyer stuff for the pre-audit, and so I was trying to figure that out. I have twice, but I wasn’t sure what buyer represented, if it represented the executive club, or if it was something else, but I can ask Dave as well.

110 00:14:11.550 ⇒ 00:14:16.330 Katherine Bayless: Here we go. Can we recheck the buyer numbers ASAP?

111 00:14:16.460 ⇒ 00:14:16.790 Katherine Bayless: Easy.

112 00:14:18.180 ⇒ 00:14:23.320 Kyle Wandel: I think we just didn’t have the twice enough data at the time we submitted it, so I have that now.

113 00:14:23.760 ⇒ 00:14:31.289 Kyle Wandel: And then, with the buyer information, I think that’s Executive Club stuff, but I’m not 100% sure. I can’t really tell from the pre-audit.

114 00:14:32.340 ⇒ 00:14:37.479 Katherine Bayless: Yeah… I don’t… it’s funny…

115 00:14:38.130 ⇒ 00:14:46.580 Katherine Bayless: I would say buyer, to me, yes, is executive clubs, but then the way Michael asked the question makes me more confused, if he says this is the twice-matching?

116 00:14:47.610 ⇒ 00:14:50.189 Kyle Wandel: Yeah, the way he’s wording the question is…

117 00:14:50.490 ⇒ 00:14:51.230 Katherine Bayless: Buyers?

118 00:14:51.230 ⇒ 00:14:51.780 Kyle Wandel: Correct.

119 00:14:52.210 ⇒ 00:14:53.480 Katherine Bayless: What? Yeah.

120 00:14:57.050 ⇒ 00:15:03.499 Katherine Bayless: Oh, okay, yeah, so Kinsey said, I’m sure it was discussed, but buyer numbers look down significantly.

121 00:15:06.630 ⇒ 00:15:12.919 Katherine Bayless: Which I think is true. I mean, I think Jackie and her team are looking at that program this year and revamping it.

122 00:15:13.170 ⇒ 00:15:18.660 Katherine Bayless: Yeah, to be totally honest, I would actually probably say let’s clarify with Michael exactly what he’s…

123 00:15:19.040 ⇒ 00:15:25.050 Katherine Bayless: asking for double-checked? Because I think the buyer’s is the exec clubs, but the way he asked the question is kind of confusing.

124 00:15:25.860 ⇒ 00:15:30.980 Kyle Wandel: Okay, I’ll reply with the twice number and then ask him for clarification on the, buyer stuff.

125 00:15:31.420 ⇒ 00:15:33.260 Katherine Bayless: Yeah. Yeah.

126 00:15:34.130 ⇒ 00:15:36.039 Katherine Bayless: That would be my recommendation.

127 00:15:36.860 ⇒ 00:15:37.690 Kyle Wandel: Alright.

128 00:15:40.040 ⇒ 00:15:46.360 Katherine Bayless: I also, I’m like… Okay, I have other questions. They’re not relevant at this time.

129 00:15:47.920 ⇒ 00:15:51.350 Katherine Bayless: Why do we edit reports in email threads?

130 00:15:52.270 ⇒ 00:15:53.810 Katherine Bayless: It’s fine, it’s fine, it’s fine.

131 00:15:59.220 ⇒ 00:16:00.060 Katherine Bayless: Anyway…

132 00:16:00.290 ⇒ 00:16:03.120 Uttam Kumaran: Yeah, so I just had, like, okay, what are the…

133 00:16:03.450 ⇒ 00:16:08.509 Uttam Kumaran: Sort of categories of usage, so this is, like, What model are people using?

134 00:16:08.930 ⇒ 00:16:13.699 Uttam Kumaran: In case that’s… I just was like, what do I have here? Let’s just throw sections together.

135 00:16:13.700 ⇒ 00:16:14.810 Katherine Bayless: Beautiful.

136 00:16:14.810 ⇒ 00:16:23.599 Uttam Kumaran: Categories for sidebar usage, I have, like, cohorts. For example, were they active in the last 7 days, 30 days, first seen?

137 00:16:23.910 ⇒ 00:16:33.399 Uttam Kumaran: Right, so, some interesting data. Tokens, I didn’t do costs yet, because I need to look again in what our token costs are, but…

138 00:16:33.930 ⇒ 00:16:35.920 Katherine Bayless: Yeah. I think this, this would just…

139 00:16:35.920 ⇒ 00:16:36.510 Uttam Kumaran: Yeah.

140 00:16:36.820 ⇒ 00:16:42.180 Katherine Bayless: I saw the email that said that the charges will start April 1st, and so.

141 00:16:42.180 ⇒ 00:16:43.060 Uttam Kumaran: Oh.

142 00:16:43.240 ⇒ 00:16:43.780 Katherine Bayless: Yeah.

143 00:16:43.780 ⇒ 00:16:45.150 Uttam Kumaran: Okay, then this is accurate.

144 00:16:45.480 ⇒ 00:16:55.920 Katherine Bayless: Yeah, yeah, so, like, I think, it’ll be, it’ll be interesting to see how much, of a bill this runs up. I still think it’s gonna be, like, comically small.

145 00:16:55.920 ⇒ 00:17:02.160 Uttam Kumaran: I don’t think it’s gonna run any… like, if I could give you a sense of scale, like, I probably…

146 00:17:02.740 ⇒ 00:17:06.030 Uttam Kumaran: Am doing, like, a billion tokens myself a month.

147 00:17:06.609 ⇒ 00:17:08.509 Uttam Kumaran: or more at Brainforge.

148 00:17:08.909 ⇒ 00:17:11.059 Uttam Kumaran: Through, like, pressure and stuff, so…

149 00:17:11.619 ⇒ 00:17:20.169 Uttam Kumaran: I don’t think… these are just big. Maybe I should just, like… yeah, these are big, but, like, kind of not really the currency, but it’s worth noting.

150 00:17:21.079 ⇒ 00:17:26.759 Katherine Bayless: Yeah, I mean, it’s funny, like, I think once the billing comes through, people will be able to see, right?

151 00:17:26.760 ⇒ 00:17:27.180 Uttam Kumaran: Yeah.

152 00:17:27.180 ⇒ 00:17:29.329 Katherine Bayless: Tokens, small billing,

153 00:17:29.330 ⇒ 00:17:37.450 Uttam Kumaran: That’s why I think it’ll… this is a good snapshot to be like, oh, you know, like, how everybody’s getting value? It’s only been, like, 50 bucks. Yeah. You know?

154 00:17:37.860 ⇒ 00:17:50.769 Katherine Bayless: Well, yeah, so this was what’s funny. I was talking to the guy from the North Carolina CPAs earlier this week, and he was asking about, you know, they were considering also doing the data share thing with remembers, and he was like, you know, what’s it costing you?

155 00:17:50.840 ⇒ 00:18:00.929 Katherine Bayless: there is the option to, like, not host your own Snowflake, but just, like, use theirs. And I was like, buddy, I’m pretty sure we’re gonna wind up paying more for the license to get the data than we will for the product.

156 00:18:00.930 ⇒ 00:18:01.580 Uttam Kumaran: Yes.

157 00:18:01.580 ⇒ 00:18:09.380 Katherine Bayless: Because our bill was $473 last month. That’s it. And that’s with all of us running full-tilt engineering on this thing.

158 00:18:09.380 ⇒ 00:18:11.369 Uttam Kumaran: Yeah… yeah. Yeah.

159 00:18:12.250 ⇒ 00:18:18.459 Uttam Kumaran: I agree. Really, like, it’s… where it’s gonna come is, like, if we have to get, like, large or XL warehouses.

160 00:18:18.780 ⇒ 00:18:19.630 Katherine Bayless: Yeah.

161 00:18:19.630 ⇒ 00:18:22.060 Uttam Kumaran: But, like, I don’t… I don’t see it…

162 00:18:22.360 ⇒ 00:18:38.900 Uttam Kumaran: I don’t see it. So, especially without the… like, really what happens is, like, I don’t know, what happened in your experience, but when I’ve used medium or large-sized warehouses is when you have, like, 100 or 200 people hitting a BI tool.

163 00:18:39.040 ⇒ 00:18:43.349 Uttam Kumaran: That’s, like, coming through a single warehouse, so then you have, like, concurrency problems.

164 00:18:43.550 ⇒ 00:18:56.839 Uttam Kumaran: And it’s usually because it’s, like, in the morning, and then, like, afternoon, like, everybody’s reports run, everybody’s this. So this is actually, like, even tackling a bigger problem, which is, like, you don’t need to have scaled warehouses.

165 00:18:57.400 ⇒ 00:18:58.410 Katherine Bayless: Like, in this…

166 00:18:58.410 ⇒ 00:19:00.139 Uttam Kumaran: Sort of ad hoc manner, you know?

167 00:19:01.710 ⇒ 00:19:08.720 Awaish Kumar: Yeah, I think I… I saw some issues recently with some of our… Like, the client’s work, like…

168 00:19:08.880 ⇒ 00:19:12.540 Awaish Kumar: It’s like, it’s not a big,

169 00:19:12.880 ⇒ 00:19:19.000 Awaish Kumar: In terms of computation, we don’t need a medium warehouse, but it’s, like, a lot of small requests at the same time.

170 00:19:19.130 ⇒ 00:19:25.419 Awaish Kumar: And it makes sense at this moment that to divide our warehouses into much more

171 00:19:25.610 ⇒ 00:19:33.979 Awaish Kumar: categories. They’re just saying, one warehouse for anybody who uses Streamlit, we can divide it by domain, like the…

172 00:19:34.250 ⇒ 00:19:39.150 Awaish Kumar: People from marketing And then one allows for people from.

173 00:19:39.150 ⇒ 00:19:46.309 Uttam Kumaran: Yes, because it’s user-based. Yeah. Yeah, it’s really hard in a BI tool to pass the warehouse, like.

174 00:19:46.860 ⇒ 00:19:52.509 Uttam Kumaran: It’s, yeah, it’s complicated to do, like, oh, this user always uses this warehouse, but in this.

175 00:19:52.740 ⇒ 00:19:58.760 Uttam Kumaran: I just change your default warehouse. We can have a Cortex warehouse, marketing Cortex warehouse, Whatever, you know?

176 00:19:59.270 ⇒ 00:20:07.960 Katherine Bayless: Yeah, I mean, it might be interesting to explore some of that, just from, like, a, like, utilization and cost kind of standpoint, but, like, I don’t think we need to go there.

177 00:20:07.960 ⇒ 00:20:09.000 Uttam Kumaran: Yeah, yeah, yeah, yeah.

178 00:20:09.000 ⇒ 00:20:10.979 Katherine Bayless: But yeah, yeah, yeah.

179 00:20:11.750 ⇒ 00:20:13.120 Katherine Bayless: Yeah, let’s go!

180 00:20:13.330 ⇒ 00:20:17.530 Awaish Kumar: More warehouses doesn’t cost anything, because it’s the usage base.

181 00:20:17.980 ⇒ 00:20:18.690 Uttam Kumaran: Yeah.

182 00:20:18.690 ⇒ 00:20:32.470 Awaish Kumar: then it’s… it’s costing us otherwise. But if we have just one warehouse for many users, so it will be, like, the response will be slower, because a lot of requests will be queued, and then it just makes it slower to respond.

183 00:20:33.080 ⇒ 00:20:33.830 Katherine Bayless: Yeah.

184 00:20:36.320 ⇒ 00:20:44.340 Uttam Kumaran: Yeah, so kind of go through tokens. I then sort of have some stuff around themes. So again, as you can see, I’m playing with every.

185 00:20:44.500 ⇒ 00:20:45.150 Katherine Bayless: sort of unique.

186 00:20:45.150 ⇒ 00:20:54.219 Uttam Kumaran: UX element that I could deal with, but, like, here are some, like, themes. Attendance, membership, it’s sort of crude.

187 00:20:54.860 ⇒ 00:20:57.000 Uttam Kumaran: Crude matching, but,

188 00:20:57.980 ⇒ 00:21:15.109 Uttam Kumaran: like, I think this is helpful to start to think about categorization, and we can go… we can go way, way deeper on this, but again, I think a great, thing, Catherine, for this team to be able, like, to share up is, like, what are the questions people are asking? Because then we can easily look at, like.

189 00:21:15.220 ⇒ 00:21:18.729 Uttam Kumaran: Could you answer this in the old world? If so, how long did it take?

190 00:21:18.850 ⇒ 00:21:21.100 Uttam Kumaran: Right? Like, I want to just really…

191 00:21:21.270 ⇒ 00:21:25.640 Uttam Kumaran: make that, like, painfully obvious, So…

192 00:21:26.160 ⇒ 00:21:30.750 Katherine Bayless: In addition, I also want to use the questions to, like, inform our engineering roadmap, right?

193 00:21:30.750 ⇒ 00:21:31.070 Uttam Kumaran: Yeah.

194 00:21:31.200 ⇒ 00:21:37.650 Katherine Bayless: once we’ve gotten more of the, like, baseline stuff done, then it’s like, okay, well, what are people asking about the most? That’s what we’ll build next, yeah.

195 00:21:37.650 ⇒ 00:21:39.040 Uttam Kumaran: Yes, yes.

196 00:21:39.980 ⇒ 00:21:42.650 Uttam Kumaran: So that’s this…

197 00:21:43.020 ⇒ 00:21:50.659 Uttam Kumaran: I didn’t find… has anyone… actually, I was supposed to do this, I didn’t actually hit the feedback. Has anyone… have you been giving feedback?

198 00:21:50.980 ⇒ 00:21:56.640 Katherine Bayless: I have not been giving it, but I did do one yesterday for you, for… I was like, oh, it’d be interesting to see if it turns up, and I.

199 00:21:56.640 ⇒ 00:21:57.130 Uttam Kumaran: Okay.

200 00:21:57.420 ⇒ 00:22:11.690 Katherine Bayless: Oh, it was, it was, like, midday yesterday, because I was working… I was kind of showing it to Christina, and I was like, ask it a question, any question, and she asked, how long has this person been on the board? And it was, like.

201 00:22:11.890 ⇒ 00:22:19.819 Katherine Bayless: what… what is this question? I don’t… like, where do you want me to… what? I don’t… I don’t know, Pat. And so I gave it a thumbs up for not guessing,

202 00:22:19.820 ⇒ 00:22:20.550 Uttam Kumaran: Okay.

203 00:22:20.740 ⇒ 00:22:21.630 Katherine Bayless: Right? Okay.

204 00:22:21.630 ⇒ 00:22:35.170 Uttam Kumaran: let me see, because I wonder… I haven’t… I didn’t use it, and yesterday, I kept procrastinating on, like, figuring… like, testing this. There’s so much other stuff, and I was like, I’ll come back and give feedback and then test the loop, but now that… if you’ve given one, then I’ll go find it.

205 00:22:35.480 ⇒ 00:22:41.720 Katherine Bayless: Yeah, you should see a thumbs, and I did put a comment in, in the feedback, so you should see both somewhere in there.

206 00:22:41.720 ⇒ 00:22:45.489 Uttam Kumaran: Okay, okay. And then, I have a little bit on, like.

207 00:22:45.630 ⇒ 00:22:50.309 Uttam Kumaran: what is in this dashboard? So, like, we don’t have stuff on, like.

208 00:22:50.420 ⇒ 00:22:57.440 Uttam Kumaran: like, it’s not in the dashboard, but there is a table for CLI usage, or Cortex agent usage.

209 00:22:57.910 ⇒ 00:23:05.750 Uttam Kumaran: again, they have, like, all these products now. Like, for example, this is Cortex Code, in SnowSight.

210 00:23:06.980 ⇒ 00:23:07.550 Uttam Kumaran: not…

211 00:23:08.050 ⇒ 00:23:18.339 Uttam Kumaran: Coco, not sidebar. So, I’ve been… I was looking for, like… I don’t know, actually, what I was looking for was, like, chat? I was actually… chat is what I was looking for, and then none of it is…

212 00:23:18.670 ⇒ 00:23:20.149 Uttam Kumaran: Has the right naming.

213 00:23:20.690 ⇒ 00:23:21.080 Katherine Bayless: Yeah.

214 00:23:21.080 ⇒ 00:23:21.650 Uttam Kumaran: Oh.

215 00:23:21.900 ⇒ 00:23:27.890 Katherine Bayless: Yeah, it is all honestly kind of funny the way it’s broken up. And I think Snowflake Intelligence is, like, the older stuff that’s gonna.

216 00:23:27.890 ⇒ 00:23:28.750 Uttam Kumaran: Yes.

217 00:23:28.750 ⇒ 00:23:29.510 Katherine Bayless: Yeah.

218 00:23:29.510 ⇒ 00:23:31.950 Uttam Kumaran: Like, document intelligence and stuff, so…

219 00:23:32.890 ⇒ 00:23:36.600 Uttam Kumaran: So that’s great, and then lastly, I have, like,

220 00:23:36.880 ⇒ 00:23:42.840 Uttam Kumaran: some, like, session-related diagnostic. This is just helpful for me to… to, like,

221 00:23:43.720 ⇒ 00:23:54.300 Uttam Kumaran: when there was an error, the error is often not very verbose, when it’s… it’ll be like, this thing is broken, and then I have, like, another section that actually, like.

222 00:23:54.440 ⇒ 00:24:04.150 Uttam Kumaran: does a verbose log of, like, what it is, so you can copy and paste it into your AI of choice and move faster.

223 00:24:04.420 ⇒ 00:24:09.599 Uttam Kumaran: So this was this, and I would love to share some of, like, the code that went into this.

224 00:24:09.830 ⇒ 00:24:10.220 Katherine Bayless: Yeah.

225 00:24:10.220 ⇒ 00:24:17.159 Uttam Kumaran: so, I have, like, sort of 3 PRs that I would love to walk this, group.

226 00:24:17.860 ⇒ 00:24:21.200 Uttam Kumaran: Through… Let me just bring this up.

227 00:24:31.420 ⇒ 00:24:38.080 Katherine Bayless: This is, like, where I’m, like, paused for appreciation. I’m like, oh, this is my first time doing PR review since my old job. It feels so nice.

228 00:24:38.080 ⇒ 00:24:43.869 Uttam Kumaran: I love VR reviews. I used to, like, when I was in office, we would just, like, actually sit next to each other.

229 00:24:43.870 ⇒ 00:24:47.149 Katherine Bayless: I like that, yeah.

230 00:24:47.150 ⇒ 00:24:51.470 Uttam Kumaran: I do too, yeah, I, it’s gotten a lot easier with AI.

231 00:24:52.100 ⇒ 00:24:56.910 Uttam Kumaran: But I still think, like, A, I can’t be like, what was the point of this whole thing, you know?

232 00:24:57.330 ⇒ 00:25:00.730 Katherine Bayless: Seriously, absolutely, yes. Yeah. Yes.

233 00:25:01.070 ⇒ 00:25:14.729 Katherine Bayless: After we’ve gone through all of the necessary business items on our agenda, I will totally share with you my latest adventures in side projects. I was talking to Kyle about them earlier. But yeah, the intent thing has been a thorn in my side.

234 00:25:15.580 ⇒ 00:25:16.770 Uttam Kumaran: Oh, really? Okay.

235 00:25:18.070 ⇒ 00:25:24.180 Uttam Kumaran: Okay, let me… Just push this…

236 00:25:33.250 ⇒ 00:25:34.060 Uttam Kumaran: Okay.

237 00:25:36.220 ⇒ 00:25:40.009 Uttam Kumaran: I just got the, like, commits there, so let’s see…

238 00:25:50.820 ⇒ 00:25:53.379 Uttam Kumaran: Alright, let’s see if the first one landed…

239 00:25:57.550 ⇒ 00:25:59.169 Uttam Kumaran: Mike, please hold.

240 00:26:01.210 ⇒ 00:26:02.560 Katherine Bayless: We’re waiting music.

241 00:26:05.570 ⇒ 00:26:09.929 Uttam Kumaran: Okay, I’m just, like, pushing this for my local. I just need to stack the branches a little bit.

242 00:26:10.430 ⇒ 00:26:13.339 Uttam Kumaran: So let me just try to finish this out.

243 00:26:30.600 ⇒ 00:26:32.120 Uttam Kumaran: Okay…

244 00:26:57.280 ⇒ 00:27:01.680 Uttam Kumaran: Sorry, something in my kit is… Slightly messed up.

245 00:27:04.860 ⇒ 00:27:10.420 Katherine Bayless: You’re good. To be totally honest, I’m suddenly realizing, my brain’s appreciating the quiet moment.

246 00:27:10.420 ⇒ 00:27:11.070 Uttam Kumaran: Okay.

247 00:27:23.700 ⇒ 00:27:24.700 Uttam Kumaran: Alright.

248 00:27:25.900 ⇒ 00:27:28.820 Uttam Kumaran: Should be all nearly there.

249 00:27:31.270 ⇒ 00:27:35.970 Uttam Kumaran: I just had, like, a bunch of work in this one branch, and I’m like, this is not good.

250 00:27:37.120 ⇒ 00:27:46.690 Uttam Kumaran: For me to ship all this in one go, so I’m just gonna split up kind of some of my work for Cortex, some of my work on the rolls, and then some of the Streamlit work.

251 00:27:46.950 ⇒ 00:27:55.189 Uttam Kumaran: But we’ll start with, we’ll start with… Yes, this one.

252 00:27:56.810 ⇒ 00:27:59.250 Uttam Kumaran: Okay, so…

253 00:28:01.700 ⇒ 00:28:03.890 Katherine Bayless: Or, like, PRA of 3.

254 00:28:03.890 ⇒ 00:28:14.260 Uttam Kumaran: Yes, so… Basically, this PR is everything around, the dbt mart for…

255 00:28:14.460 ⇒ 00:28:18.770 Uttam Kumaran: like, adoption. I kind of went a little bit further and just made sure we had

256 00:28:19.210 ⇒ 00:28:25.170 Uttam Kumaran: kind of, like, as much as we needed just for the future reporting, like, and I kind of went through…

257 00:28:25.280 ⇒ 00:28:29.750 Uttam Kumaran: All the documentation, tried to bring up all of the… like…

258 00:28:30.100 ⇒ 00:28:41.980 Uttam Kumaran: for example, like, if you look at feedback events, there’s things like, when was it resolved? Like, I don’t… we’re… I don’t… I’m not clear yet whether we will be using these functionality, and, like, even how to do it, entirely, but…

259 00:28:42.220 ⇒ 00:28:47.290 Uttam Kumaran: I, like, didn’t want to, like, leave… leave something out, and then be like, where is this later?

260 00:28:47.430 ⇒ 00:28:50.350 Uttam Kumaran: But the rough structure here…

261 00:28:50.660 ⇒ 00:28:58.290 Uttam Kumaran: And maybe what I can do is… I can just… Go into this branch.

262 00:29:06.200 ⇒ 00:29:08.410 Uttam Kumaran: Oh, Cortex PRA.

263 00:29:17.860 ⇒ 00:29:24.160 Uttam Kumaran: So we basically are left with, like, this core mart, everything, around…

264 00:29:24.430 ⇒ 00:29:33.130 Uttam Kumaran: Each of the products, and then all the things that you saw, which is, like, what models people are using, when are they hitting requests.

265 00:29:33.320 ⇒ 00:29:37.199 Uttam Kumaran: Query themes, feedbacks, and then some…

266 00:29:37.430 ⇒ 00:29:42.710 Uttam Kumaran: some consolidated reports, like daily metrics, for example.

267 00:29:43.160 ⇒ 00:29:54.640 Uttam Kumaran: So, like, what… we’re kind of working our way backwards. The way I try to have the mart set up is that we have a clear sense of, like, the users, clear sense of, like, the feedback that’s coming back.

268 00:29:54.790 ⇒ 00:30:04.040 Uttam Kumaran: And then we have a clear sense of, like, everything around tokens, and then we sort of arrange some of these reporting models. So, really, like, if you go to the… if we go to the int…

269 00:30:04.230 ⇒ 00:30:13.890 Uttam Kumaran: intermediate section, of Cortex adoption. Like, if we were to look at user query events, you’re gonna see that we’re just bringing in

270 00:30:14.650 ⇒ 00:30:21.629 Uttam Kumaran: kind of, like, from the raw… from the raw, like, this is sort of what the raw table is named. We’re just bringing in

271 00:30:21.920 ⇒ 00:30:34.310 Uttam Kumaran: exactly the query that was run, the SQL that was run, and details about when it was run. And then if you look at tokens, we’re bringing in all the token-related information. There’s some,

272 00:30:34.830 ⇒ 00:30:38.570 Uttam Kumaran: like, flattening and some expansion I need to do to get, like.

273 00:30:38.830 ⇒ 00:30:45.170 Uttam Kumaran: they just throw everything into a JSON for the token-related information, so I pulled some of that out.

274 00:30:45.440 ⇒ 00:30:48.830 Uttam Kumaran: So this should hopefully solve, like.

275 00:30:49.210 ⇒ 00:30:53.500 Uttam Kumaran: the model that gets used and the tokens. I don’t think as of now.

276 00:30:53.740 ⇒ 00:30:59.199 Uttam Kumaran: any of our users are going to be really switching. I also have not checked whether we can fix that, like, fix it to one.

277 00:31:01.770 ⇒ 00:31:06.579 Uttam Kumaran: Maybe it’s worth… Maybe it’s worth me doing that, just so, like, again, as little…

278 00:31:07.090 ⇒ 00:31:12.829 Uttam Kumaran: need for the people to kind of make decisions as possible, I don’t know. Or I don’t know if you’ve actually seen any, like, difference…

279 00:31:13.950 ⇒ 00:31:29.869 Katherine Bayless: I… yeah, it’s funny, I… I don’t think I have any bripes with just leaving it on auto, honestly. I think where I would start to have more of an opinion will be when, like, inevitably, you know, there’s some sort of, like, crappy model they throw in there, and then it, like, leans on that one because it’s cheaper.

280 00:31:29.870 ⇒ 00:31:30.500 Uttam Kumaran: Yes.

281 00:31:30.500 ⇒ 00:31:32.459 Katherine Bayless: But I feel like perplexity does that sometimes.

282 00:31:32.460 ⇒ 00:31:34.389 Uttam Kumaran: Yes, Perplexity does. I always switch it.

283 00:31:34.820 ⇒ 00:31:43.460 Katherine Bayless: Yeah. So, Cortex code hasn’t, hasn’t, hurt me, hurt my feelings at all yet, but, yeah, I think Auto is okay.

284 00:31:44.950 ⇒ 00:31:47.540 Katherine Bayless: So yeah, and then we’re doing… Did you switch it, though?

285 00:31:47.540 ⇒ 00:31:54.420 Uttam Kumaran: Yeah, so then we’re doing some cleanup on, like, the actual events itself. Right now, we don’t have, like.

286 00:31:56.060 ⇒ 00:32:04.350 Uttam Kumaran: notion of, like, a team or a role, and these aren’t, like, the, this isn’t, like, access roles. So, the way I…

287 00:32:04.620 ⇒ 00:32:09.170 Uttam Kumaran: Would like to ask this crew is, like, we can tag users by team.

288 00:32:09.370 ⇒ 00:32:17.330 Uttam Kumaran: I just need to think about… we just need to think about, like, sort of the taxonomy. So, that was a question I had in the chat, which was.

289 00:32:17.850 ⇒ 00:32:25.219 Uttam Kumaran: Do we have a clear understanding of, like, is there separate teams?

290 00:32:25.340 ⇒ 00:32:36.369 Uttam Kumaran: at this moment, do they require any, like, actual functional access changes? Or… I would actually just like to tag them as such so that I can start to show adoption by team.

291 00:32:37.100 ⇒ 00:32:40.990 Uttam Kumaran: We don’t necessarily need to do roll, we can just assume…

292 00:32:41.570 ⇒ 00:32:46.799 Uttam Kumaran: Same rule, but I don’t know, I feel like team may be helpful, but I… there isn’t a notion of groups.

293 00:32:47.110 ⇒ 00:32:50.729 Uttam Kumaran: within Snowflake, so I will have to use the metadata tags.

294 00:32:51.730 ⇒ 00:32:58.219 Katherine Bayless: Gotcha. Yeah, so I think… At present, we are, operating under the

295 00:32:58.550 ⇒ 00:33:14.310 Katherine Bayless: things that we have should be visible to all the folks. However, yeah, to your point about future-proofing, I think capturing the team, and then… I mean, user role, honestly, it might almost make more sense to capture, and I mean, I’m so not a hierarchy person, but, like.

296 00:33:14.380 ⇒ 00:33:31.229 Katherine Bayless: probably that one ought to be, like, title, rank, level kind of thing, just because I can imagine future states where we are bringing in data that is more, sort of, like, politically sensitive, and it might be, like, VPs and above can access this, or, you know, only the finance team leaders.

297 00:33:31.230 ⇒ 00:33:32.050 Uttam Kumaran: Exactly.

298 00:33:32.190 ⇒ 00:33:32.850 Uttam Kumaran: I don’t…

299 00:33:32.850 ⇒ 00:33:34.989 Katherine Bayless: want those situations to happen, but…

300 00:33:34.990 ⇒ 00:33:38.209 Uttam Kumaran: Finance Executive is the typically two common ones.

301 00:33:38.520 ⇒ 00:33:46.639 Uttam Kumaran: And then what we will do is we will actually, we will… so we will create a functional… we will create a snowflake role for that.

302 00:33:46.760 ⇒ 00:33:52.979 Uttam Kumaran: And that will be… Like, something that that group of people has access to.

303 00:33:54.940 ⇒ 00:34:01.330 Uttam Kumaran: And then, yeah, there may be, like, actually also just, like, levels of access, but ultimately, for me, it’s just…

304 00:34:01.710 ⇒ 00:34:04.630 Uttam Kumaran: this is where Snowflake doesn’t have any, like.

305 00:34:04.870 ⇒ 00:34:14.870 Uttam Kumaran: further than just, like, roles, and, like, you have access to a role to your user, and then maybe you have, like, certain roles can access other roles. There’s no notion of groups.

306 00:34:15.040 ⇒ 00:34:19.840 Uttam Kumaran: So, I can use Teams as the metadata, and then…

307 00:34:20.170 ⇒ 00:34:24.239 Uttam Kumaran: So yeah, I mean, one thing I can do is, like, if it’s helpful for me to just be like, here’s everybody.

308 00:34:24.780 ⇒ 00:34:27.319 Uttam Kumaran: Tell me their teams, I could do it that way.

309 00:34:27.320 ⇒ 00:34:33.060 Katherine Bayless: Well, actually, Can we pass it with SCIM? Like, since the users are getting provisioned via Okta?

310 00:34:33.060 ⇒ 00:34:42.869 Uttam Kumaran: So that’s why I didn’t know the Trellica Okta piece. Yeah, actually, that would be preferred. If you… I don’t know if… if we’re… if we’re ready for that conversation, I would love to basically…

311 00:34:42.980 ⇒ 00:34:46.359 Uttam Kumaran: Tell them what, what, I mean, ideally…

312 00:34:46.620 ⇒ 00:34:53.850 Uttam Kumaran: ultimately, we’re telling them to map the create user statement, right? So I can tell them which properties to map to.

313 00:34:54.560 ⇒ 00:34:58.360 Uttam Kumaran: What, on their side? Yeah, I… ideally, and then, yeah.

314 00:34:58.690 ⇒ 00:35:11.439 Katherine Bayless: So, Ian and I, we started looking at it on Tuesday, I think it was, and he’s… whenever we’re ready, he’s ready, to set up the rest of the, like, pieces that we wanted. We did learn, though, that…

315 00:35:12.180 ⇒ 00:35:14.820 Katherine Bayless: It sounds like there might be some…

316 00:35:14.950 ⇒ 00:35:27.360 Katherine Bayless: I don’t think it matters, but there’s potentially benefit to having Ian define the roles in Okta, and so then on the first, like, assignment, they push them into Snowflake.

317 00:35:27.360 ⇒ 00:35:30.810 Uttam Kumaran: Find the role in terms of, oh, like, when…

318 00:35:30.810 ⇒ 00:35:33.250 Katherine Bayless: Literally just the role name, actually.

319 00:35:35.570 ⇒ 00:35:37.509 Uttam Kumaran: Yeah, I mean, I was thinking, like.

320 00:35:37.680 ⇒ 00:35:40.120 Uttam Kumaran: For example, like, the data team…

321 00:35:40.820 ⇒ 00:35:45.539 Uttam Kumaran: we are all role developer. There’s also, like, role analyst.

322 00:35:46.670 ⇒ 00:35:52.560 Uttam Kumaran: I just don’t know… I think what I want to avoid is, like, okay, now, if we make a role change.

323 00:35:53.100 ⇒ 00:35:56.929 Uttam Kumaran: I guess it doesn’t matter, because it’s mainly on creation, so…

324 00:35:58.610 ⇒ 00:36:03.860 Katherine Bayless: Yeah, and he was not saying that we needed to do it that way, it just, it seemed like there might be.

325 00:36:03.860 ⇒ 00:36:06.980 Uttam Kumaran: I would prefer for him to do default warehouse, default role.

326 00:36:07.150 ⇒ 00:36:08.330 Katherine Bayless: Yeah, yeah, yeah.

327 00:36:08.330 ⇒ 00:36:19.099 Uttam Kumaran: And I can tell them to do that. Yeah, I have that already, at least for, like, everybody. And so, this is also where, like, let’s say somebody new joins the data team, I can just change it after they join.

328 00:36:19.280 ⇒ 00:36:22.570 Uttam Kumaran: Just so everybody gets… everybody can get, like.

329 00:36:23.110 ⇒ 00:36:27.179 Uttam Kumaran: what we said, which is read on Prod Martz and Cortex.

330 00:36:27.300 ⇒ 00:36:32.199 Uttam Kumaran: And then I have a default warehouse for everybody.

331 00:36:32.440 ⇒ 00:36:37.469 Uttam Kumaran: And then that’s their default role, and then if I can get their first, last email.

332 00:36:37.790 ⇒ 00:36:40.250 Uttam Kumaran: And then I would love… I have some metadata asks.

333 00:36:40.520 ⇒ 00:36:42.139 Uttam Kumaran: That’s perfect. That’s it.

334 00:36:42.890 ⇒ 00:36:49.850 Katherine Bayless: Okay, okay, yeah, I… honestly, if you wanna, like, just throw together the wishlist, I’m sure Ian would knock it out real fast. He’s awesome.

335 00:36:49.850 ⇒ 00:36:51.320 Uttam Kumaran: Okay, okay, great.

336 00:36:51.970 ⇒ 00:36:55.050 Katherine Bayless: He might be on our channel, actually, I can’t remember, let me check.

337 00:36:55.300 ⇒ 00:36:56.720 Uttam Kumaran: Okay, great, yeah, I can just…

338 00:36:56.960 ⇒ 00:37:03.870 Uttam Kumaran: Yeah, and I would love to kind of… I’m just curious, like, even, like, what it looks like on his site in terms of the Octa mapping, or the… or Trollica, yeah.

339 00:37:04.780 ⇒ 00:37:07.650 Katherine Bayless: Yeah, okay, he’s not in the channel, but we can add him. Yes.

340 00:37:07.650 ⇒ 00:37:08.130 Uttam Kumaran: No problem.

341 00:37:08.130 ⇒ 00:37:22.919 Katherine Bayless: But yeah, actually, he did also say that there is the option of, like, they would still be, obviously, because the SSO, the user would be provisioned in Okta, but, like, the role and stuff could be governed in Trellica versus Okta and stuff. Yeah, I was like, this is all over.

342 00:37:22.920 ⇒ 00:37:29.320 Uttam Kumaran: I would prefer that they… they run the default role in warehouse, because we can change it, and then there’s no, like…

343 00:37:29.670 ⇒ 00:37:31.379 Uttam Kumaran: There’s no two-way sync.

344 00:37:31.560 ⇒ 00:37:43.409 Uttam Kumaran: So, if someone joins and then needs a higher level, I’d actually prefer that. One, so they don’t land with public. That’s, like, so there’s, like, the first provision, it’s kind of, like, off our hands.

345 00:37:43.440 ⇒ 00:37:44.420 Katherine Bayless: And then…

346 00:37:44.420 ⇒ 00:37:47.749 Uttam Kumaran: There’s at least, like, a check to get higher level access.

347 00:37:48.000 ⇒ 00:37:48.800 Uttam Kumaran: Right.

348 00:37:49.270 ⇒ 00:37:55.699 Katherine Bayless: Yeah, yeah, yeah, exactly, exactly. I’m with you, yeah, and I think Ian’s like, whatever makes your life easier.

349 00:37:56.810 ⇒ 00:37:58.260 Uttam Kumaran: Okay, cool.

350 00:37:58.490 ⇒ 00:38:03.409 Uttam Kumaran: And then… On the staging side for Cortex.

351 00:38:03.410 ⇒ 00:38:04.340 Katherine Bayless: Like…

352 00:38:04.340 ⇒ 00:38:06.800 Uttam Kumaran: I’ll just kind of give you a peek of, like.

353 00:38:07.280 ⇒ 00:38:15.390 Uttam Kumaran: This is how we’re trying to break stuff up from, like, There’s a view called… this.

354 00:38:15.710 ⇒ 00:38:26.399 Uttam Kumaran: that I, like, do some cleanup on, and then basically peel back all the feedbacks. But this is… this is, I think, maybe broken, so I need to go check this out. But,

355 00:38:27.620 ⇒ 00:38:34.459 Uttam Kumaran: like, oh, I just clicked on the same one, like, let me find… I think this is actually, like, the first one, yeah.

356 00:38:34.930 ⇒ 00:38:38.670 Uttam Kumaran: So this is, like… the first…

357 00:38:39.290 ⇒ 00:38:44.689 Uttam Kumaran: the, like, there’s no raw, right, because we’re not importing anything, so this is the first view that gets created.

358 00:38:44.900 ⇒ 00:38:51.430 Uttam Kumaran: Where I’m, like, pulling out all of these items from Cortex analyst requests.

359 00:38:51.640 ⇒ 00:38:52.520 Uttam Kumaran: view.

360 00:38:53.060 ⇒ 00:38:54.260 Uttam Kumaran: Gotcha.

361 00:38:54.520 ⇒ 00:38:58.620 Uttam Kumaran: So that’s why, if you’re… if you’re interested in going to the source on any of these, I would…

362 00:38:59.140 ⇒ 00:39:03.670 Uttam Kumaran: Totally recommend, Poking in this,

363 00:39:04.120 ⇒ 00:39:08.459 Uttam Kumaran: in the staging cortex adoption, because I’ve done a lot of that heavy lifting here.

364 00:39:10.560 ⇒ 00:39:13.620 Uttam Kumaran: Cool, so that’s everything on…

365 00:39:13.900 ⇒ 00:39:16.700 Uttam Kumaran: sort of, like, the marts, like.

366 00:39:17.160 ⇒ 00:39:24.989 Uttam Kumaran: I didn’t expect there to be a ton of questions there, but let me know if there’s any questions. I… yeah, I think this one is…

367 00:39:25.650 ⇒ 00:39:29.550 Uttam Kumaran: Mainly just, like, can we get everything into some tables for us to look at?

368 00:39:30.270 ⇒ 00:39:31.319 Uttam Kumaran: On the adoption side.

369 00:39:31.320 ⇒ 00:39:38.869 Katherine Bayless: really awesome. Like, I genuinely, I’m very excited to test my scale-must-be-to-our-advantage hypothesis and start using the interactive.

370 00:39:38.870 ⇒ 00:39:39.510 Uttam Kumaran: Yes.

371 00:39:39.750 ⇒ 00:39:40.889 Katherine Bayless: the development.

372 00:39:41.290 ⇒ 00:39:50.080 Uttam Kumaran: Yeah, and I’m really hopeful that, like, we have a nice flywheel with the dashboard, so this’ll all run on our schedule,

373 00:39:50.320 ⇒ 00:39:58.000 Uttam Kumaran: And so that’s what I’m… I’m pumped, I think, kind of, like, have a bow on this piece. And then let me show about,

374 00:39:58.450 ⇒ 00:39:59.859 Uttam Kumaran: the dashboard.

375 00:40:00.000 ⇒ 00:40:07.190 Uttam Kumaran: So… Let me see what’s the best way for me to, like…

376 00:40:09.130 ⇒ 00:40:16.899 Uttam Kumaran: I did… okay, let’s, let me show you just, like, what… the script is. So,

377 00:40:17.680 ⇒ 00:40:22.089 Uttam Kumaran: Let’s see… So, right now, in… in,

378 00:40:23.520 ⇒ 00:40:26.110 Uttam Kumaran: Can I just… not see all the diffs?

379 00:40:26.580 ⇒ 00:40:32.790 Uttam Kumaran: So, in this branch, I have, like, our current version of streamlitapp.py.

380 00:40:32.900 ⇒ 00:40:45.640 Uttam Kumaran: Dreamlit runs on these PY files, it’s just, like, one long Python file. You can do external imports, it’s just gonna take some, like, UI configuring for me to allow Snowflake to… to…

381 00:40:45.810 ⇒ 00:40:53.229 Uttam Kumaran: Import some non, standard libraries, like, math and stuff is, like, normal.

382 00:40:53.500 ⇒ 00:40:59.190 Uttam Kumaran: But I will… I’m gonna go explore some of the… fancier…

383 00:40:59.360 ⇒ 00:41:01.790 Uttam Kumaran: viz libraries to see what we can bring in.

384 00:41:02.280 ⇒ 00:41:08.969 Uttam Kumaran: I don’t think there… I think this, I used AI for quite a bit of it. I mean, I remember having to, like, write

385 00:41:09.210 ⇒ 00:41:16.950 Uttam Kumaran: these, like, a few years ago. But really, I think my bigger point on this was actually, like, all about,

386 00:41:17.680 ⇒ 00:41:20.239 Uttam Kumaran: How we do the deployment step.

387 00:41:20.460 ⇒ 00:41:36.219 Uttam Kumaran: And so I wrote this, I wrote this… script, which is… Here, somewhere… Let’s see…

388 00:41:37.180 ⇒ 00:41:38.300 Uttam Kumaran: Yes.

389 00:41:40.010 ⇒ 00:41:44.180 Uttam Kumaran: So, this is sort of actually, like…

390 00:41:44.650 ⇒ 00:41:51.849 Uttam Kumaran: It’s only 30 lines, but it’s actually, like, kind of the… what is allowing us to go execute,

391 00:41:51.970 ⇒ 00:42:03.079 Uttam Kumaran: a replacement of the Streamlit app. So, one of the problems I faced initially was you can’t create a new Streamlit app from the CLI. So that is a manual step.

392 00:42:03.200 ⇒ 00:42:17.919 Uttam Kumaran: That I found. I may… I may still be wrong about that, but I just wasn’t able to get it to work. But what I found was it’s actually totally possible for you to,

393 00:42:18.910 ⇒ 00:42:30.399 Uttam Kumaran: deploy a local app to Snowflake. So what… what I kind of… what I kind of do is… is you’re able to make a bunch of changes, and then this script basically connects

394 00:42:30.500 ⇒ 00:42:39.260 Uttam Kumaran: do Streamlit, replaces the app, and then you can just refresh. And so, all the README on how this works is,

395 00:42:39.410 ⇒ 00:42:47.720 Uttam Kumaran: In here… So that, in case you’re using Claude Code or something else, there is,

396 00:42:48.280 ⇒ 00:42:53.840 Uttam Kumaran: you shouldn’t have to, like, mess around too much, as long as your credentials are configured.

397 00:42:54.160 ⇒ 00:42:59.800 Uttam Kumaran: I don’t… I think, like, yeah, it’s really, like, in this, like, Part B, which is just connecting to… to…

398 00:43:00.200 ⇒ 00:43:05.499 Uttam Kumaran: the Snow SQL, and then… deploying, like, the Streamwood apps.

399 00:43:06.950 ⇒ 00:43:10.699 Katherine Bayless: Do you think tomorrow, on the call, we could, like, meet.

400 00:43:10.700 ⇒ 00:43:12.419 Uttam Kumaran: Yeah, I could do, like, a setup.

401 00:43:12.620 ⇒ 00:43:18.100 Katherine Bayless: Yeah, like, I would love to get this up and running for all of us. Okay, cool. Yeah. Yeah.

402 00:43:18.100 ⇒ 00:43:23.870 Uttam Kumaran: Yeah, so maybe… maybe I’ll merge all this, and then I’ll just send a note, which is just, like,

403 00:43:24.640 ⇒ 00:43:28.170 Uttam Kumaran: like, this is sort of… yeah, I’ll send a note on just, like.

404 00:43:28.680 ⇒ 00:43:36.430 Uttam Kumaran: make sure your thing is updated for main. I don’t think… I want to just make sure that everything is working for everybody, in particular, like.

405 00:43:37.030 ⇒ 00:43:44.769 Uttam Kumaran: You will just… it’ll most likely fail on, like, credentials, so just making sure that you have your credentials saved locally, and that you can,

406 00:43:45.530 ⇒ 00:43:51.479 Uttam Kumaran: run through a test where you make an edit, to a SnowSite dash, streamline dashboard, and then go for it.

407 00:43:52.090 ⇒ 00:43:55.789 Uttam Kumaran: So that’s this…

408 00:43:55.790 ⇒ 00:43:56.620 Katherine Bayless: It’ll make our…

409 00:43:56.620 ⇒ 00:43:56.950 Uttam Kumaran: Yeah.

410 00:43:56.950 ⇒ 00:44:09.759 Katherine Bayless: development, like, iteration loops way faster with membership, like, if we’re able to leverage, like, all the things in code right there with the AI to help, and then, like, you know, membership wants to add job titles to the primary representative, push, right?

411 00:44:09.760 ⇒ 00:44:18.969 Uttam Kumaran: Yes, yeah, so that was exactly it. I mean, I was doing a lot of Streamlit and Snowflake stuff, like, maybe a year ago, and it was so brutal.

412 00:44:19.460 ⇒ 00:44:24.749 Uttam Kumaran: And because you can’t, like, replicate the environment locally that easily.

413 00:44:25.110 ⇒ 00:44:26.400 Katherine Bayless: Right, right.

414 00:44:26.400 ⇒ 00:44:37.689 Uttam Kumaran: I’m like, this sucks. So, I was like, I’m frustrated. I need to figure this out. So I think I got it working. So yeah, let’s totally do… let’s totally do that tomorrow.

415 00:44:37.850 ⇒ 00:44:44.130 Uttam Kumaran: Okay. And then the last peak piece I… I, did, like.

416 00:44:44.620 ⇒ 00:44:52.460 Uttam Kumaran: do some changes on… on grants, as well as I added some GitHub actions that will verify access.

417 00:44:52.650 ⇒ 00:44:55.890 Uttam Kumaran: One of the… so I’ll talk about the first piece.

418 00:44:56.290 ⇒ 00:45:02.860 Uttam Kumaran: Let’s see… So, I created a role,

419 00:45:03.210 ⇒ 00:45:07.739 Uttam Kumaran: developer. Actually, this is another verification step.

420 00:45:08.150 ⇒ 00:45:10.959 Uttam Kumaran: Yeah, so this is really the one I want to look at.

421 00:45:11.740 ⇒ 00:45:16.430 Uttam Kumaran: So I created a role developer, and…

422 00:45:16.770 ⇒ 00:45:22.150 Uttam Kumaran: I created, like, I was just, like, going back and forth on, like, how to document, and I thought this was really great, which basically…

423 00:45:22.640 ⇒ 00:45:26.540 Uttam Kumaran: it, like… I was just trying to, like, be… say if you were gonna say.

424 00:45:26.700 ⇒ 00:45:33.639 Uttam Kumaran: hey, like, how did… why did we come up with this role? It’s pretty clear to say, like, what does this role have? So, role developer.

425 00:45:33.780 ⇒ 00:45:40.140 Uttam Kumaran: You know, basically has… Roll DevRite, Fraud Marts,

426 00:45:40.360 ⇒ 00:45:43.420 Uttam Kumaran: Staging, and so you can actually see

427 00:45:43.590 ⇒ 00:45:48.679 Uttam Kumaran: That, like, hey, this is the CTA, canonical Edge Default.

428 00:45:48.810 ⇒ 00:45:57.499 Uttam Kumaran: brings in role dev right, as well as Role Chat, and then we have the grants here to role developer.

429 00:45:58.840 ⇒ 00:46:08.530 Uttam Kumaran: Currently, we don’t… we’re not allowing staging or marts, because staging was just done by the CI. I mean, marts is just also by the GitHub action.

430 00:46:09.370 ⇒ 00:46:11.880 Uttam Kumaran: we can change that. I think, for the most part.

431 00:46:12.000 ⇒ 00:46:19.360 Uttam Kumaran: we haven’t needed to, like, urgently edit anything in staging, and I think our pre-R process is working.

432 00:46:19.470 ⇒ 00:46:26.370 Uttam Kumaran: Like, I think, Kyle, you’ve been able to look at the code and then check it out in dev, and I’ve been able to edit in dev and then push to staging, so…

433 00:46:27.030 ⇒ 00:46:38.080 Uttam Kumaran: this is sort of the safest, and so this is sort of, I think, the kind of go forward. And then I also created additional roles, so that’s just, like, the role developer, and then

434 00:46:38.370 ⇒ 00:46:44.890 Uttam Kumaran: We also have a role for… Yeah, Cortex Analyst…

435 00:46:45.500 ⇒ 00:46:50.280 Uttam Kumaran: Yes, that’s that one. Let me just make sure… Yes, okay.

436 00:46:51.030 ⇒ 00:46:56.050 Uttam Kumaran: So this is Cortex Analyst… User.

437 00:46:56.590 ⇒ 00:47:02.029 Uttam Kumaran: So… this is another fun thing I had to figure out, which was, like.

438 00:47:02.380 ⇒ 00:47:04.090 Uttam Kumaran: Oh, how do you, like…

439 00:47:04.840 ⇒ 00:47:16.170 Uttam Kumaran: I’m sure, Catherine, you’re noticing that streamlets live in, like, a schema, and then, like, the, assets you have, you’re allowed to use are, like.

440 00:47:17.020 ⇒ 00:47:20.819 Uttam Kumaran: Not global, so there’s this, like, database role.

441 00:47:20.960 ⇒ 00:47:26.229 Uttam Kumaran: called Cortex Analyst User, which allows you to see the sidebar.

442 00:47:28.260 ⇒ 00:47:28.730 Katherine Bayless: Okay.

443 00:47:28.730 ⇒ 00:47:29.460 Uttam Kumaran: And…

444 00:47:29.460 ⇒ 00:47:29.900 Katherine Bayless: Okay.

445 00:47:29.900 ⇒ 00:47:31.630 Uttam Kumaran: Like, as you can see, it’s like…

446 00:47:31.930 ⇒ 00:47:35.269 Uttam Kumaran: I have to… it’s not… it’s like a snowflake,

447 00:47:35.710 ⇒ 00:47:39.569 Uttam Kumaran: system role. Yeah. Which is, like.

448 00:47:40.260 ⇒ 00:47:43.619 Uttam Kumaran: I feel like this is, like, lazy engineering on their part, but…

449 00:47:43.920 ⇒ 00:47:48.189 Uttam Kumaran: fine. So I had to… I first, like, was like, okay, like, how do I…

450 00:47:48.520 ⇒ 00:47:50.830 Uttam Kumaran: Toggle this on and off, and then so…

451 00:47:51.140 ⇒ 00:47:57.689 Uttam Kumaran: I’m… I’m adding it to, role DevWrite and Role Developer.

452 00:47:57.900 ⇒ 00:48:05.499 Uttam Kumaran: And then we… we then roll a version of this up into the Role Cortex Analyst. So…

453 00:48:05.800 ⇒ 00:48:09.200 Uttam Kumaran: This is, like, a little bit of a unique…

454 00:48:09.340 ⇒ 00:48:17.259 Uttam Kumaran: piece about, like, the Cortex Analyst user, and then the one last piece… that I’ll share is,

455 00:48:17.630 ⇒ 00:48:24.260 Uttam Kumaran: This is, like, sort of my baseline for, like, the role chat user. So…

456 00:48:24.960 ⇒ 00:48:29.989 Uttam Kumaran: This is just, like, the grant that I ran, which is, like, okay, all these people have role chat.

457 00:48:30.270 ⇒ 00:48:34.650 Uttam Kumaran: And then I sat there, default,

458 00:48:34.780 ⇒ 00:48:41.579 Uttam Kumaran: warehouse and their default, role to role chat, warehouse cortex.

459 00:48:43.380 ⇒ 00:48:44.950 Uttam Kumaran: And then that…

460 00:48:45.120 ⇒ 00:48:52.209 Uttam Kumaran: that previous policy rolls into role chat. So that way, role chat just has Fraud Marts and the chat access.

461 00:48:53.930 ⇒ 00:48:58.910 Uttam Kumaran: So, hopefully this, like, allows us… now I think we have, like, two core users. We have, like, us.

462 00:48:59.250 ⇒ 00:49:03.220 Uttam Kumaran: plus J, and then we have,

463 00:49:04.090 ⇒ 00:49:07.630 Uttam Kumaran: just the people in who can access chat and read Prod Marts.

464 00:49:08.370 ⇒ 00:49:08.920 Katherine Bayless: Yep.

465 00:49:09.630 ⇒ 00:49:16.489 Uttam Kumaran: I… I think the biggest thing for us to kind of think about is on the naming side, is just… we want these to reflect

466 00:49:17.090 ⇒ 00:49:23.449 Uttam Kumaran: Like, what you can do, like, as we talked about, it should be, like, a functional name.

467 00:49:23.570 ⇒ 00:49:32.179 Uttam Kumaran: I was thinking about roll chat read, but there is no role chat, like, write, so I was like, okay, we’ll just go with roll chat, and then roll prod Marts.

468 00:49:32.620 ⇒ 00:49:33.600 Uttam Kumaran: Read.

469 00:49:33.770 ⇒ 00:49:37.029 Uttam Kumaran: So then, if we have a third user.

470 00:49:38.490 ⇒ 00:49:44.009 Uttam Kumaran: I guess we have to think about, like, in what situation, what other access they would need, but we can…

471 00:49:44.720 ⇒ 00:49:46.620 Uttam Kumaran: come up with that, I think that’s fine.

472 00:49:46.830 ⇒ 00:49:53.460 Katherine Bayless: I think the only other one that we have is the one that you, I think, had already made, right? The, like, the Streamlit Creator one.

473 00:49:53.460 ⇒ 00:49:54.910 Uttam Kumaran: Oh, yeah, that’s correct, yeah.

474 00:49:54.910 ⇒ 00:49:58.940 Katherine Bayless: Yeah, yeah, so it’s like, the only thing they can write is a Streamlit app.

475 00:49:59.190 ⇒ 00:50:08.980 Uttam Kumaran: Yeah, okay, fair. So yeah, Broadmart’s read will allow you to look at Streamlit, but Streamlit Creator Yeah, so…

476 00:50:10.180 ⇒ 00:50:18.620 Uttam Kumaran: I’m ho- we’re hopefully, like, very close on, like, all the roll stuff. Still just, like, making sure, and I want to put a visual into the repo so it’s, like, really clear, but…

477 00:50:18.850 ⇒ 00:50:22.439 Uttam Kumaran: I guess my question was gonna be if I can…

478 00:50:23.700 ⇒ 00:50:29.130 Uttam Kumaran: maybe, like, start to deprecate. We have… we have a couple of other roles in there.

479 00:50:29.360 ⇒ 00:50:33.159 Uttam Kumaran: Maybe I can even just check what they are.

480 00:50:33.270 ⇒ 00:50:39.099 Uttam Kumaran: that I was like, okay, maybe we’re ready to deprecate once I just, like, run a confirmation that everybody can…

481 00:50:39.220 ⇒ 00:50:40.730 Uttam Kumaran: Retain access.

482 00:50:43.190 ⇒ 00:50:43.630 Katherine Bayless: Yeah.

483 00:50:43.630 ⇒ 00:50:44.019 Uttam Kumaran: I see.

484 00:50:44.020 ⇒ 00:50:48.140 Katherine Bayless: You can also remove the SDG folks from the grants.

485 00:50:48.140 ⇒ 00:50:48.680 Uttam Kumaran: Okay.

486 00:50:48.860 ⇒ 00:50:54.440 Katherine Bayless: Because, like, I mean, we will potentially work with them again later this year, but we don’t need to give them access in the…

487 00:50:54.580 ⇒ 00:50:56.319 Katherine Bayless: Okay. Yeah.

488 00:50:56.320 ⇒ 00:50:56.910 Uttam Kumaran: Okay.

489 00:50:57.140 ⇒ 00:50:59.070 Katherine Bayless: Just leave… not leaving these bends around.

490 00:50:59.180 ⇒ 00:51:00.160 Katherine Bayless: Okay.

491 00:51:00.410 ⇒ 00:51:10.340 Uttam Kumaran: So, like, a couple of roles, we have, like, Cortex Demo Restricted, Early Adopter, And then… Snowflake analysts.

492 00:51:11.050 ⇒ 00:51:14.560 Uttam Kumaran: I may just confirm and consolidate.

493 00:51:15.220 ⇒ 00:51:15.940 Katherine Bayless: Or, like…

494 00:51:15.940 ⇒ 00:51:23.639 Uttam Kumaran: Yeah, I may just check and see how I can just consolidate some of those, if that’s okay. We also have Okta Provisioner, but I assume that’s, like, the Okta… okay.

495 00:51:23.870 ⇒ 00:51:36.850 Katherine Bayless: Yeah. You can totally, consolidate to your heart’s content. I very much cede this control. I think the Snowflake Analyst one, I think, is the one that occurs.

496 00:51:36.850 ⇒ 00:51:37.660 Uttam Kumaran: Oh.

497 00:51:37.660 ⇒ 00:51:47.909 Katherine Bayless: being set, but Ian, if we’ll get him, you know, we’ll tell him which ones we want it to set, and then we can clean that one up. Okay. But yeah, I think Snowflake Analyst is what it’s currently sending.

498 00:51:48.220 ⇒ 00:51:48.880 Uttam Kumaran: Okay.

499 00:51:49.160 ⇒ 00:51:55.500 Uttam Kumaran: And then, really, like, this is where I’ve just been like, okay, like, how do I… I want to start… I haven’t tested this,

500 00:51:55.820 ⇒ 00:52:01.300 Uttam Kumaran: I haven’t tested this just yet, but I wanted to create a GitHub action that verified Access.

501 00:52:01.410 ⇒ 00:52:06.550 Uttam Kumaran: So, I have… I have to… GitHub Actions a little bit finicky, but I want to test this, which is basically, like.

502 00:52:06.730 ⇒ 00:52:10.450 Uttam Kumaran: It just…

503 00:52:11.220 ⇒ 00:52:25.260 Uttam Kumaran: it just runs a check using, like, a certain role, and then this, script verify role developer access is here, and basically, tries to, like, run a…

504 00:52:25.870 ⇒ 00:52:28.169 Uttam Kumaran: It tries to run a… it tries… basically, like.

505 00:52:28.920 ⇒ 00:52:32.769 Uttam Kumaran: Looks… tries to run a couple of queries to… to just check whether

506 00:52:33.100 ⇒ 00:52:46.920 Uttam Kumaran: anything that we’re pushing messes up a role. Additionally, I want to always just try to… this is something I’ve… this is something, like, net new, I don’t think we’ve ever done this, but I want to try to just enforce, like, the people that we expect to have certain access always have certain access.

507 00:52:46.920 ⇒ 00:52:47.570 Katherine Bayless: Hmm.

508 00:52:48.170 ⇒ 00:52:53.769 Uttam Kumaran: So I was like, alright, I feel like this is a good time to just think about, like, a little action that can execute.

509 00:52:55.280 ⇒ 00:52:56.270 Katherine Bayless: I like that.

510 00:52:57.140 ⇒ 00:53:04.770 Uttam Kumaran: That can execute this, so… We’ll see if it works, and hopefully never figures.

511 00:53:04.890 ⇒ 00:53:13.660 Uttam Kumaran: Or actually, I mean, maybe I should see if it actually works, and then… Yes, it’s not triggering, because it’s actually… we’re in a good spot, but… I just thought, like.

512 00:53:14.230 ⇒ 00:53:20.869 Uttam Kumaran: sometimes, like, these role changes can be, like, finicky, so I just want to make sure that anytime we’re pushing a new PR,

513 00:53:21.000 ⇒ 00:53:24.760 Uttam Kumaran: Like, we can verify that that access is in the right place, things like that, so…

514 00:53:25.250 ⇒ 00:53:35.029 Katherine Bayless: Yeah, no, I think that’s very smart. Because even, actually, like, yesterday, it was funny, while you guys were making some of the changes, Anna Rutter on the membership team reached out to me and Kyle.

515 00:53:35.030 ⇒ 00:53:35.730 Uttam Kumaran: Yeah.

516 00:53:35.730 ⇒ 00:53:43.930 Katherine Bayless: streamlit anymore! And it was like, okay, don’t panic. I mean, we’re totally fine, it was just funny, so yeah, like… Good, that’s good!

517 00:53:44.380 ⇒ 00:53:46.780 Uttam Kumaran: People are using stuff. That’s great.

518 00:53:47.290 ⇒ 00:53:59.750 Katherine Bayless: So one question about roles. So, I know it’s not necessarily the best habit that I’ve gotten into, but I’m gonna pretend that it’s okay. But so, like, the remembers, like, the data share, I mean, it’s so big.

519 00:53:59.750 ⇒ 00:54:00.190 Uttam Kumaran: Yeah.

520 00:54:00.190 ⇒ 00:54:06.439 Katherine Bayless: stuff in there, and we’re working with the membership team right now to do those journeys, and so what I’ve been doing is, like.

521 00:54:06.440 ⇒ 00:54:22.919 Katherine Bayless: using Coco to basically build some, like, temporary views under the raw schema, where I’m like, okay, this is the logic, this is the code, this is the SQL we ultimately need to run through and create the dbt models around. But it just gives me, like, a way to kind of, like.

522 00:54:23.220 ⇒ 00:54:37.400 Katherine Bayless: build something, to get, things moving for this journey stuff, so, like, I don’t know if for a role developer, we should give it access to the data share, or if that should be a different role. I’m okay with it being a different role that I, like, hop into when I’m doing this.

523 00:54:37.630 ⇒ 00:54:38.850 Katherine Bayless: But I…

524 00:54:38.850 ⇒ 00:54:43.790 Uttam Kumaran: This is what I ended up running, so I… it was, again, it’s another type of privilege.

525 00:54:43.790 ⇒ 00:54:45.569 Katherine Bayless: Yeah, yeah, yeah.

526 00:54:45.570 ⇒ 00:54:50.439 Uttam Kumaran: So, I can just… I think I ran this, I don’t know if… have you tried it today? I can just double check.

527 00:54:50.750 ⇒ 00:55:05.660 Katherine Bayless: Actually, I haven’t had a chance to try it today. Kyle was working with it earlier, and potentially the way that it responded to a question made us wonder if it wasn’t able to see the data share. Okay. But maybe, maybe it was just actually getting the question correct.

528 00:55:05.660 ⇒ 00:55:18.390 Kyle Wandel: Well, I mean, I think it was just pulling… the difference was, so, like, the… what we were talking about was Anna was able to do a query, and it was pulling… pulling from dev, so she… we changed that, so she can’t do that anymore. But then, by doing that.

529 00:55:19.000 ⇒ 00:55:41.659 Kyle Wandel: So, it pulled that information from dev. Today, when I ran the query, basically the same query, it was pulling from prod, and I was… and I was thinking that because it’s pulling from prod, it would go to that prod environment first, and not the remember’s environment. I don’t know if that’s how it would work, because I don’t know if it’s that smart, but that was my only…

530 00:55:41.730 ⇒ 00:55:42.370 Kyle Wandel: Thinking.

531 00:55:42.370 ⇒ 00:55:45.680 Uttam Kumaran: Yeah, I guess now I’m understanding. Yeah, I guess…

532 00:55:48.330 ⇒ 00:55:52.050 Uttam Kumaran: Yeah, I guess because I would say it’s sort of, like.

533 00:55:52.370 ⇒ 00:55:55.569 Uttam Kumaran: Doesn’t follow the pattern, maybe we create another role.

534 00:55:56.090 ⇒ 00:56:00.940 Uttam Kumaran: That way, It’s… yeah.

535 00:56:01.110 ⇒ 00:56:07.109 Uttam Kumaran: Yeah, I think maybe we do that. Maybe we create another… we create… I’ll create another role called,

536 00:56:07.830 ⇒ 00:56:10.670 Uttam Kumaran: I guess roll, data share read or something?

537 00:56:10.930 ⇒ 00:56:26.599 Katherine Bayless: Yeah, actually, do that, because I think then, for the membership team, like, we could tell them, like, you can also use this role if you do want to go poke around in the raw data to see… because, like, so, totally different scenario.

538 00:56:26.600 ⇒ 00:56:41.909 Katherine Bayless: But the other person on the membership team, adorably, stopped by my office yesterday and was like, I thought of you the other day, I had to export data from Remembers, and it crashed my computer when I tried to open the Excel, because it was 200,000 rows, and I was like, it’s in Snowflake, why didn’t you just go there? And she’s like.

539 00:56:43.230 ⇒ 00:56:56.759 Katherine Bayless: Yeah. You know, right? So, like, I think there are probably also scenarios where I would want them to be able to get into the raw data, so, like, yeah, I like the idea of data share read as a dedicated role, and then we have to consciously choose to behave that way.

540 00:56:57.220 ⇒ 00:57:02.539 Uttam Kumaran: Okay, so then I’ll create role data share read. People, for the most part, should have role chat.

541 00:57:03.280 ⇒ 00:57:06.670 Uttam Kumaran: And we will have… Sort of all.

542 00:57:07.090 ⇒ 00:57:07.680 Katherine Bayless: Yeah.

543 00:57:07.850 ⇒ 00:57:08.970 Katherine Bayless: Superpowers.

544 00:57:09.390 ⇒ 00:57:09.770 Kyle Wandel: Super powerful.

545 00:57:09.770 ⇒ 00:57:11.100 Uttam Kumaran: Ours, yeah.

546 00:57:11.670 ⇒ 00:57:32.220 Kyle Wandel: The one thing that doesn’t need a solution right this second, but it would be good to think about soon, and Catherine and I talked about this, is this idea of, like, permissions on certain metrics. I know we can do it by table or, like, column, but there are certain metrics that people just shouldn’t have access to, so, like, financial information for exhibitor data, but we would love for them

547 00:57:32.260 ⇒ 00:57:43.619 Kyle Wandel: to be able to access the idea of how much they saved on their exhibitor application or registration. So they still kind of need access to that column.

548 00:57:43.790 ⇒ 00:57:55.900 Kyle Wandel: But they don’t… I mean, it’s probably… you have to basically have to create a new report, and then limit them to only that report, is my guess. But we may need to do some role provisioning based on that level, and not just.

549 00:57:57.570 ⇒ 00:57:58.540 Uttam Kumaran: The whole marts.

550 00:57:59.390 ⇒ 00:58:09.580 Katherine Bayless: Yeah, so I think… I’m hoping we can… similar to the, like, you know, getting into, like, super granular permissioning around teams, I’m also hoping we can avoid this, but…

551 00:58:10.140 ⇒ 00:58:28.949 Katherine Bayless: that I could see some of the, you know, leadership fear winding up, sort of, getting into a situation where it’s like, they don’t want us to be able to, like… or they wouldn’t want certain members of certain teams to see, like, raw financials, but they could acknowledge that there are use cases for, like, the aggregate information coming out of that data.

552 00:58:29.080 ⇒ 00:58:36.959 Katherine Bayless: I don’t think we need to make steps that direction yet, because I’m really hoping we can avoid it, but Kyle is right that it could come down the road.

553 00:58:36.960 ⇒ 00:58:38.660 Uttam Kumaran: There’s some, yeah, there’s some solves.

554 00:58:39.800 ⇒ 00:58:40.440 Katherine Bayless: Yeah.

555 00:58:40.440 ⇒ 00:58:46.949 Uttam Kumaran: I will… there’s some solves, I will not… we don’t have to go into it, but yes, I think there’s some… there’s some ways to solve.

556 00:58:47.300 ⇒ 00:58:48.599 Katherine Bayless: Yeah, yeah.

557 00:58:48.710 ⇒ 00:58:49.570 Katherine Bayless: Yeah.

558 00:58:49.570 ⇒ 00:59:09.279 Katherine Bayless: It’s like, it’s one of those things, and it’s funny, even Christina was, like, you know, totally on board with my, stance of, like, it’s good that the engineering folks know that there’s a lot of potential in the platform for restrictions in clever ways, but, yeah, don’t make that sound like it’s the, you know, the thing we want everybody to start thinking about setting up, right? You’re like, yeah, right.

559 00:59:09.280 ⇒ 00:59:10.170 Uttam Kumaran: Yeah, yeah.

560 00:59:10.170 ⇒ 00:59:11.980 Katherine Bayless: Yeah.

561 00:59:11.980 ⇒ 00:59:12.510 Uttam Kumaran: Makes sense.

562 00:59:12.510 ⇒ 00:59:13.240 Katherine Bayless: Somebody asks.

563 00:59:13.760 ⇒ 00:59:15.290 Katherine Bayless: Right? But…

564 00:59:15.430 ⇒ 00:59:28.030 Katherine Bayless: On that note, I do have a touch base with Christina tomorrow, and then I have, lunch with Kinsey on Monday, and so I can report back as to the, go-no-go, go with modifications, status at that point.

565 00:59:28.630 ⇒ 00:59:29.140 Uttam Kumaran: Okay.

566 00:59:29.940 ⇒ 00:59:30.610 Uttam Kumaran: G’day.

567 00:59:30.990 ⇒ 00:59:32.199 Katherine Bayless: I’m very optimistic.

568 00:59:32.780 ⇒ 00:59:33.520 Uttam Kumaran: Yeah.

569 00:59:35.340 ⇒ 00:59:39.930 Katherine Bayless: Yeah, I think this all looks really great, like… Yeah.

570 00:59:40.490 ⇒ 00:59:50.440 Uttam Kumaran: Yeah, so I wanna kind of… so I think I’m gonna leave the stream lit where it is, I’m gonna push this model. I am going to switch to go work on chat improvements.

571 00:59:52.170 ⇒ 01:00:10.639 Uttam Kumaran: So, I felt like this took longer than I expected to get the adoption metrics going, but they didn’t make it really easy. So I think I’ll do that. I think I’m happy with at least having something in Streamlit, so I’m just gonna make sure all the pipes are connected, I’ll merge some of this stuff today, and then I’m…

572 01:00:10.700 ⇒ 01:00:12.259 Uttam Kumaran: I think, ideally…

573 01:00:12.950 ⇒ 01:00:19.390 Uttam Kumaran: maybe we have, like, again, if I can… if I… if Kai can even start taking over and editing of the adoption dashboard.

574 01:00:19.390 ⇒ 01:00:19.830 Katherine Bayless: And, like.

575 01:00:19.830 ⇒ 01:00:21.929 Uttam Kumaran: She’s able to edit it, perfect.

576 01:00:22.350 ⇒ 01:00:29.530 Uttam Kumaran: And then one thing I can take on is just, like, biz improvements and things like that, but I want to move to…

577 01:00:30.110 ⇒ 01:00:34.279 Uttam Kumaran: Doing a lot more work on the actual context building for… for Coco.

578 01:00:34.820 ⇒ 01:00:47.620 Katherine Bayless: Yeah, yeah. Oh, and actually, this is another, funny thing Kyle and I were thinking about earlier. It might be worth also thinking about, like, forbidden questions, like, so.

579 01:00:47.620 ⇒ 01:00:48.340 Uttam Kumaran: Okay.

580 01:00:48.340 ⇒ 01:00:57.949 Katherine Bayless: right now, somebody could go in and be like, you know, I need to downsize this team. Which team members have produced the least sales, or something like that, which is kind of a

581 01:00:58.130 ⇒ 01:01:12.029 Katherine Bayless: solid question, but not the kind of organization we are. And so yeah, like, the idea of, like, an anti-question might be an interesting thing to explore. Like, don’t let people ask which of their colleagues appear laziest based on data.

582 01:01:12.220 ⇒ 01:01:23.609 Uttam Kumaran: Yes, yes. Yeah, I… I think… okay, that’s actually a really great point, so I will maybe do a ticket on, like, yeah, topics to not talk about, and…

583 01:01:24.360 ⇒ 01:01:28.159 Uttam Kumaran: It’s almost the opposite of, like… yeah, it’s like a different type of steering, yeah.

584 01:01:28.160 ⇒ 01:01:29.860 Katherine Bayless: Okay. Yeah.

585 01:01:29.990 ⇒ 01:01:36.770 Katherine Bayless: It was just funny, we were like, oh yeah, we can almost, like, come up with, like, a list of, like, you know, bad, gossipy-type questions. Yeah.

586 01:01:36.770 ⇒ 01:01:37.940 Uttam Kumaran: That’s a good point.

587 01:01:39.960 ⇒ 01:01:46.270 Uttam Kumaran: But yeah, well, yeah, okay, that’s actually helpful, too. Okay.

588 01:01:46.380 ⇒ 01:01:48.049 Katherine Bayless: Cool. So that’s kind of like…

589 01:01:48.200 ⇒ 01:01:57.829 Uttam Kumaran: Yeah, yeah, very interesting. Okay, so I think let’s plan on maybe doing, like, a streamlit editing walkthrough tomorrow.

590 01:01:58.240 ⇒ 01:01:59.130 Uttam Kumaran: and then…

591 01:01:59.260 ⇒ 01:02:09.599 Uttam Kumaran: Yeah, I think, Awash, if you want to close out anything remaining on the ID stitching, I think that’d be… that’d be awesome, and I know you’re still waiting for a couple reviews, so I feel like…

592 01:02:10.080 ⇒ 01:02:12.299 Uttam Kumaran: Pretty productive week, you know, overall.

593 01:02:13.260 ⇒ 01:02:15.830 Katherine Bayless: Yeah, yeah, yeah, no, I think this is awesome.

594 01:02:15.830 ⇒ 01:02:24.340 Katherine Bayless: I would love to, next week, so we’re also doing this work with ProServe, right, for on those AWS accounts.

595 01:02:24.340 ⇒ 01:02:37.020 Katherine Bayless: So we did our three discovery calls today, and then next week we’re gonna do the, I don’t know, launch-ready plan. Some of the project management aspects of ProServe just drive me crazy. I’m like, I feel like we just meet to talk about meetings.

596 01:02:37.030 ⇒ 01:02:45.319 Katherine Bayless: But that’s fine. But we’re also gonna do a workshop, I think it’s either gonna be Wednesday or Thursday, around, like.

597 01:02:45.560 ⇒ 01:02:46.570 Katherine Bayless: just, like…

598 01:02:46.640 ⇒ 01:02:56.770 Katherine Bayless: I’m not as familiar with some of the latest and greatest in AWS services around AI, right? Like, I mean, Bedrock was probably the last thing I went deep on learning, and now there’s all the quick, sweet stuff.

599 01:02:56.780 ⇒ 01:03:12.220 Katherine Bayless: And so, if you guys would like to just, like, you know, listen in or be in that conversation to sort of, like, listen for things where it’s like, oh, that actually might make a lot of sense to put some effort into provisioning, you’re welcome to, not obligated. Okay.

600 01:03:12.630 ⇒ 01:03:13.249 Katherine Bayless: Welcome to you.

601 01:03:14.310 ⇒ 01:03:15.020 Uttam Kumaran: Okay.

602 01:03:18.940 ⇒ 01:03:19.710 Uttam Kumaran: Okay.

603 01:03:19.870 ⇒ 01:03:20.880 Katherine Bayless: That’s all I got.

604 01:03:21.630 ⇒ 01:03:22.210 Uttam Kumaran: Cool.

605 01:03:23.060 ⇒ 01:03:28.429 Uttam Kumaran: Alright, well, hope everyone has a nice rest of the day, and talk to you tomorrow morning.

606 01:03:28.780 ⇒ 01:03:30.809 Katherine Bayless: Okay, cool. Thanks, guys.

607 01:03:31.380 ⇒ 01:03:31.990 Kyle Wandel: There you go.

608 01:03:31.990 ⇒ 01:03:32.570 Uttam Kumaran: Bye.

Brainforge Knowledge

Explorer

2026-03-26_cta_x_brainforge_weekly_sync_dfad8681

Graph View