Meeting Title: AI Team | Planning ／ Retro Date: 2025-04-01 Meeting participants: Uttam Kumaran, Amber Lin, Miguel De Veyra, Casie Aviles
WEBVTT
1 00:03:15.090 ⇒ 00:03:17.180 Miguel de Veyra: Assalam Alayku, my friend
2 00:03:26.650 ⇒ 00:03:27.960 Amber Lin: Hi team.
3 00:03:29.090 ⇒ 00:03:30.000 Casie Aviles: Hey! Amber.
4 00:03:30.670 ⇒ 00:03:35.060 Amber Lin: Hello! I’ll keep this quick. I don’t know if your time’s morning
5 00:03:35.060 ⇒ 00:03:39.370 Miguel de Veyra: He’s coming, he’s told he’s gonna be running late
6 00:03:39.900 ⇒ 00:03:42.720 Amber Lin: Oh, okay, sounds good.
7 00:03:43.290 ⇒ 00:03:46.830 Amber Lin: Let’s see if he has checked the tickets yet.
8 00:03:46.830 ⇒ 00:03:47.860 Miguel de Veyra: No, no! I moved it back
9 00:03:49.490 ⇒ 00:03:50.310 Amber Lin: No.
10 00:03:50.310 ⇒ 00:03:52.209 Miguel de Veyra: I moved the tickets back
11 00:03:52.820 ⇒ 00:03:54.210 Amber Lin: Oh, I see!
12 00:03:54.210 ⇒ 00:03:56.380 Miguel de Veyra: Yeah. Let me share my screen.
13 00:03:56.610 ⇒ 00:03:57.300 Amber Lin: Okay.
14 00:03:58.870 ⇒ 00:04:00.180 Miguel de Veyra: Yeah, so
15 00:04:00.690 ⇒ 00:04:07.479 Miguel de Veyra: yeah, so basically cause as Uta mentioned, right? If anyone in the company should be able to do the tickets.
16 00:04:07.740 ⇒ 00:04:10.099 Miguel de Veyra: So I I need to be basically
17 00:04:10.100 ⇒ 00:04:11.440 Amber Lin: Wow!
18 00:04:11.440 ⇒ 00:04:13.609 Miguel de Veyra: It has to be a bit more detailed. Yeah.
19 00:04:14.150 ⇒ 00:04:17.250 Amber Lin: That’s so so detailed, Miguel
20 00:04:17.640 ⇒ 00:04:18.440 Miguel de Veyra: Yeah.
21 00:04:19.390 ⇒ 00:04:27.759 Miguel de Veyra: I was working on this earlier. But yeah, so for example, the way it’s gonna work is we’re gonna have basically
22 00:04:30.720 ⇒ 00:04:36.019 Miguel de Veyra: 2 ways to do it. So for example, for Zoom Meetings and stuff, we’re gonna store them into S 3,
23 00:04:36.260 ⇒ 00:04:48.159 Miguel de Veyra: because S. 3 is for files, basically videos, images, anything of the sort. And then, for example, for I believe for slack messages, I I asked them a lot
24 00:04:48.310 ⇒ 00:04:55.859 Miguel de Veyra: here. Yeah, basically, we want, we don’t need to store images or text messages into
25 00:04:56.400 ⇒ 00:04:58.540 Amber Lin: Textual context into
26 00:04:59.790 ⇒ 00:05:06.960 Miguel de Veyra: Basically s, 3 anymore. Because what we can do is just put them into super base, like what we’re doing now, Casey, and then just embed them
27 00:05:08.730 ⇒ 00:05:10.950 Miguel de Veyra: So I don’t think it should be a problem
28 00:05:10.950 ⇒ 00:05:12.320 Amber Lin: Is, for
29 00:05:13.040 ⇒ 00:05:21.120 Miguel de Veyra: Oh, yeah, super base. So what’s gonna happen is, I probably need to document this somewhere. So what’s gonna happen is, wait. Let’s
30 00:05:23.020 ⇒ 00:05:25.350 Miguel de Veyra: wait. Let’s do it for
31 00:05:28.930 ⇒ 00:05:34.560 Miguel de Veyra: let’s just add it here, page in front page info.
32 00:05:36.540 ⇒ 00:05:43.159 Miguel de Veyra: So let’s just say, super base for vector embeddings.
33 00:05:45.200 ⇒ 00:05:51.510 Miguel de Veyra: And then s, 3 s. For large files.
34 00:05:55.010 ⇒ 00:06:05.150 Miguel de Veyra: And then what’s the other one? Snowflake snowflake is for real that so embeddings
35 00:06:06.750 ⇒ 00:06:10.599 Miguel de Veyra: large files and then dashboards.
36 00:06:11.664 ⇒ 00:06:13.939 Miguel de Veyra: So this is how it’s going to be
37 00:06:15.368 ⇒ 00:06:19.890 Miguel de Veyra: but yeah, I’m I just clarified that like a few moments ago. So I’m going to clean this up
38 00:06:20.100 ⇒ 00:06:25.419 Miguel de Veyra: because Uttam, as Utham mentioned yesterday. Right, the fields should also be here.
39 00:06:25.620 ⇒ 00:06:27.140 Miguel de Veyra: The you know.
40 00:06:28.200 ⇒ 00:06:28.960 Amber Lin: Oh! Oh!
41 00:06:28.960 ⇒ 00:06:33.739 Miguel de Veyra: It’s gonna be in snowflake and stuff. But yeah, that should be. That should be okay.
42 00:06:34.160 ⇒ 00:06:40.190 Amber Lin: Okay, so how is how are we? Gonna what are we gonna work on this week?
43 00:06:44.310 ⇒ 00:06:50.050 Miguel de Veyra: It’s gonna be, it’s gonna be once utham does this stuff bye
44 00:06:50.050 ⇒ 00:06:59.660 Miguel de Veyra: like he moves ready for development it’s gonna be, I I’m guessing it’s gonna be mostly this Zoom Meetings, I think, takes priority and slack messages.
45 00:07:00.790 ⇒ 00:07:04.279 Miguel de Veyra: Then I want to move one for Jana so she can start working on something
46 00:07:04.500 ⇒ 00:07:05.350 Amber Lin: I see.
47 00:07:05.940 ⇒ 00:07:08.499 Miguel de Veyra: Basically, all of this are like priorities.
48 00:07:08.870 ⇒ 00:07:10.439 Miguel de Veyra: I need to work on them
49 00:07:11.038 ⇒ 00:07:17.169 Amber Lin: What about today? Oh, today, you guys are off perfect. No need for today. Then
50 00:07:17.170 ⇒ 00:07:17.530 Miguel de Veyra: Yeah, okay.
51 00:07:17.530 ⇒ 00:07:18.520 Amber Lin: Us.
52 00:07:19.510 ⇒ 00:07:21.550 Casie Aviles: I’m sorry I just have a question.
53 00:07:22.130 ⇒ 00:07:26.090 Casie Aviles: So wait is, is the Zoom Meeting part still?
54 00:07:26.800 ⇒ 00:07:34.920 Casie Aviles: I mean, do we? Is it still in the the core. I mean, it’s the requirement still in review, because I started working on it already. And I,
55 00:07:36.520 ⇒ 00:07:40.949 Casie Aviles: yeah, I finished the 1st automation just to transfer
56 00:07:41.350 ⇒ 00:07:46.420 Casie Aviles: the incoming meetings to S. 3 already. So it’s deployed
57 00:07:47.510 ⇒ 00:07:49.629 Miguel de Veyra: Where are you storing it in? S. 3. By the way.
58 00:07:50.646 ⇒ 00:07:59.830 Casie Aviles: I saw a I think, a folder. I’m not sure if they’re called folders. But yeah, it’s like a folder there called internal AI, I think, and then Zoom Meetings.
59 00:08:00.617 ⇒ 00:08:02.160 Miguel de Veyra: Yeah, that’s why. Wait. Let me
60 00:08:02.580 ⇒ 00:08:03.369 Casie Aviles: Yeah, you’re good.
61 00:08:04.780 ⇒ 00:08:05.480 Casie Aviles: Oh.
62 00:08:05.880 ⇒ 00:08:06.825 Miguel de Veyra: What is this?
63 00:08:08.180 ⇒ 00:08:11.119 Miguel de Veyra: I think? Wait! Sorry, guys. I am not sure.
64 00:08:11.790 ⇒ 00:08:14.630 Miguel de Veyra: Let me just go here. Go here.
65 00:08:16.390 ⇒ 00:08:20.289 Miguel de Veyra: I don’t know how to log in here. I’m not gonna like
66 00:08:22.080 ⇒ 00:08:23.710 Amber Lin: Is it? Do you want to share your screen? Then
67 00:08:24.080 ⇒ 00:08:25.069 Casie Aviles: Yeah, of course.
68 00:08:26.200 ⇒ 00:08:31.100 Miguel de Veyra: Okay, yeah, because we’re gonna restructure that
69 00:08:31.210 ⇒ 00:08:33.970 Miguel de Veyra: aws, I structured it like that.
70 00:08:34.450 ⇒ 00:08:38.140 Miguel de Veyra: But then Melody and I discussed that it’s gonna be per client
71 00:08:38.840 ⇒ 00:08:45.750 Casie Aviles: I see. Yeah, I mean, yeah, there’s really no structure. I guess it’s just a very straightforward
72 00:08:46.893 ⇒ 00:08:50.450 Casie Aviles: structure right now that I did. It’s the same thing I did with
73 00:08:51.080 ⇒ 00:08:57.749 Casie Aviles: Google. So I have all these meetings as you can see here. So I started getting since yesterday
74 00:08:57.990 ⇒ 00:09:03.480 Miguel de Veyra: Yeah. So ideally, we wanna, for example, for Zoom Meetings, there’s gonna be client.
75 00:09:03.620 ⇒ 00:09:04.990 Miguel de Veyra: And then
76 00:09:05.330 ⇒ 00:09:18.980 Miguel de Veyra: there’s gonna be, I believe, for example, ABC, and then the Avi coffee. And then for ABC, there’s gonna be Zoom Meetings for ABC and then slack messages, emails and everything else. Get up for ABC.
77 00:09:19.910 ⇒ 00:09:20.580 Casie Aviles: Okay.
78 00:09:21.180 ⇒ 00:09:26.750 Miguel de Veyra: So we’re gonna restructure. That is this, did you do this via anything
79 00:09:27.680 ⇒ 00:09:34.819 Casie Aviles: No, no, I use the Api. So I I have the scripts on windmill, actually. So it’s a bunch of python scripts
80 00:09:35.210 ⇒ 00:09:36.179 Miguel de Veyra: Okay. Okay.
81 00:09:36.180 ⇒ 00:09:42.520 Casie Aviles: So I, yeah, I basically moved away from the no code automation tools that we have. So even Zapier, I
82 00:09:43.327 ⇒ 00:09:46.800 Casie Aviles: ditch that already. So I just used the code, since
83 00:09:46.940 ⇒ 00:09:49.110 Casie Aviles: I want to have it in one place
84 00:09:49.500 ⇒ 00:09:51.069 Miguel de Veyra: Okay, yeah, that makes sense
85 00:09:51.810 ⇒ 00:09:54.649 Casie Aviles: But yeah, yeah, this is what it looks like right now.
86 00:09:56.950 ⇒ 00:10:03.250 Miguel de Veyra: Okay. Okay, okay, yeah, I’m gonna that’s fine for now, I think.
87 00:10:03.430 ⇒ 00:10:08.550 Miguel de Veyra: Cause we’re gonna at least structure. At least, you know, we don’t have to start from scratch, but we’ll probably have to work on it again.
88 00:10:10.097 ⇒ 00:10:19.920 Miguel de Veyra: Yeah. Cause as, yeah, there was like a lot of changes yesterday, like we didn’t know that we basically, Utm has to review the technical requirements before we can proceed.
89 00:10:20.200 ⇒ 00:10:22.029 Miguel de Veyra: I think that was fairly new.
90 00:10:22.300 ⇒ 00:10:24.619 Miguel de Veyra: I just needed yesterday. So yeah.
91 00:10:29.720 ⇒ 00:10:32.790 Casie Aviles: Yeah, I guess just one more thing that I
92 00:10:33.558 ⇒ 00:10:44.890 Casie Aviles: have to work on is, since we already have a lot of you know, existing videos. That one. I haven’t really transferred those yet. So that’s going to be a different, I guess a different thing altogether.
93 00:10:45.380 ⇒ 00:10:57.909 Miguel de Veyra: Yeah. And then for the existing videos is, did we target like, anyway? Because the way I’m thinking about it is we have to kinda separate. That is, that like a sales call right with
94 00:10:58.560 ⇒ 00:10:59.210 Miguel de Veyra: internal
95 00:10:59.570 ⇒ 00:11:09.369 Casie Aviles: Well, we we don’t really have metadata. So that’s that part I didn’t implement yet. But what I did have was, you know, just a very
96 00:11:10.060 ⇒ 00:11:17.840 Casie Aviles: a simple solution like separating them in specific folder categories, like, you know, AI team data team, you know.
97 00:11:18.810 ⇒ 00:11:20.549 Miguel de Veyra: Is it also per client or no?
98 00:11:22.340 ⇒ 00:11:27.509 Casie Aviles: I I also created like a folder for yeah, I could just share again
99 00:11:27.510 ⇒ 00:11:28.330 Miguel de Veyra: Okay.
100 00:11:29.870 ⇒ 00:11:33.780 Casie Aviles: So we have our recordings here. I I know it’s very messy right now. But
101 00:11:34.250 ⇒ 00:11:39.360 Casie Aviles: and yeah, the automation doesn’t really categorize it. But I have these folders
102 00:11:39.740 ⇒ 00:11:40.200 Miguel de Veyra: Yes.
103 00:11:40.200 ⇒ 00:11:48.299 Casie Aviles: So ideally, we, we place the recording in the correct one. So we have clients. So ABC, and this. So
104 00:11:48.640 ⇒ 00:11:53.809 Casie Aviles: we have some recordings here. So that’s how I structured it initially. But
105 00:11:54.490 ⇒ 00:12:00.510 Miguel de Veyra: Yes, I think that one we just do the same as on your
106 00:12:00.820 ⇒ 00:12:02.900 Miguel de Veyra: or can you create folders in S 3
107 00:12:04.415 ⇒ 00:12:05.050 Casie Aviles: Yeah.
108 00:12:05.540 ⇒ 00:12:06.750 Miguel de Veyra: That’s what I did.
109 00:12:07.530 ⇒ 00:12:12.780 Miguel de Veyra: Okay, okay, are we using the same? S. 3. We are right. Rainforge internal. I
110 00:12:14.310 ⇒ 00:12:20.530 Miguel de Veyra: wait. Sorry. Let me double check. Why can’t I log in S. 3
111 00:12:39.183 ⇒ 00:12:40.959 Casie Aviles: I’m not. I’m not sure why. But
112 00:12:41.260 ⇒ 00:12:43.869 Casie Aviles: this is where I access. S. 3.
113 00:12:45.410 ⇒ 00:12:48.210 Casie Aviles: This there’s this aws access portal
114 00:12:48.210 ⇒ 00:12:49.480 Miguel de Veyra: What’s the bucket name?
115 00:12:50.210 ⇒ 00:12:55.999 Casie Aviles: Oh, the bucket name! It’s internal AI bucket. Then Zoom Meetings
116 00:12:56.000 ⇒ 00:13:02.840 Miguel de Veyra: Oh, yeah, yeah. Okay, okay, were you the one who created this? Because I don’t. I think, yeah. Asia Pacific. I created this. Okay? Okay, yeah.
117 00:13:03.920 ⇒ 00:13:06.810 Miguel de Veyra: You created this. Right? I saw this already.
118 00:13:07.690 ⇒ 00:13:09.280 Casie Aviles: Yeah, I didn’t make this
119 00:13:09.620 ⇒ 00:13:10.610 Miguel de Veyra: Okay, okay.
120 00:13:10.790 ⇒ 00:13:13.529 Miguel de Veyra: And then, did you add any folders here?
121 00:13:14.030 ⇒ 00:13:14.990 Casie Aviles: Yeah, these are basically
122 00:13:15.281 ⇒ 00:13:19.360 Miguel de Veyra: Okay, okay. The the root folders that you didn’t add, okay. I was confused
123 00:13:19.760 ⇒ 00:13:23.690 Casie Aviles: Yes, yes, and then there’s just the files here
124 00:13:34.180 ⇒ 00:13:35.379 Miguel de Veyra: Yeah, I think.
125 00:13:42.280 ⇒ 00:13:45.499 Casie Aviles: Yeah, that’s pretty much what I worked on internally.
126 00:13:51.670 ⇒ 00:13:58.519 Amber Lin: Let me see like, how’s the I know we have also a ticket s. 3 setup. That’s done right.
127 00:13:58.520 ⇒ 00:14:00.039 Miguel de Veyra: Yeah, this is basically it
128 00:14:00.390 ⇒ 00:14:05.509 Amber Lin: Oh, I see great. Oh, cause we escalated and done a lot of helped out, I remember.
129 00:14:07.288 ⇒ 00:14:13.210 Amber Lin: Let’s see, yeah. So what is our goal for the end of this week because Uton wants
130 00:14:13.340 ⇒ 00:14:20.119 Amber Lin: progress. And how are we gonna show that to him like, what are we gonna get done this week?
131 00:14:21.880 ⇒ 00:14:24.390 Miguel de Veyra: I mean the S. 3 is done already.
132 00:14:24.500 ⇒ 00:14:27.299 Miguel de Veyra: I I mean the test. The setup is done so
133 00:14:27.760 ⇒ 00:14:30.750 Miguel de Veyra: Then, I believe, let’s focus on
134 00:14:31.775 ⇒ 00:14:34.469 Miguel de Veyra: the slack messages and the Zoom Meetings
135 00:14:35.480 ⇒ 00:14:35.960 Amber Lin: Okay.
136 00:14:35.960 ⇒ 00:14:37.690 Casie Aviles: I think that’s reasonable.
137 00:14:37.690 ⇒ 00:14:44.169 Miguel de Veyra: Yeah, those 2 are the only things I would say we focus on while Jan is working on what was assigned to Jana. Sorry
138 00:14:44.590 ⇒ 00:14:45.550 Miguel de Veyra: the
139 00:14:45.550 ⇒ 00:14:47.560 Casie Aviles: And emails in there
140 00:14:47.560 ⇒ 00:14:51.899 Miguel de Veyra: Yeah. But yeah, let’s just focus on this, too. For now, cause I’ve mapped out majority of this.
141 00:14:52.080 ⇒ 00:15:00.629 Miguel de Veyra: the thing, the only thing I need to add there, Casey is because S. 3 isn’t really a storage, for it’s not really a database. It’s a file storage right
142 00:15:01.530 ⇒ 00:15:02.620 Casie Aviles: Yes, yes.
143 00:15:02.620 ⇒ 00:15:12.940 Miguel de Veyra: So we basically still need to put these Zoom Meetings into super base or snowflake. I would. I don’t think it should be in. Do we need to run rag on this meetings?
144 00:15:14.150 ⇒ 00:15:16.200 Casie Aviles: I think that was one of the
145 00:15:16.200 ⇒ 00:15:17.409 Miguel de Veyra: This is not the idea
146 00:15:17.410 ⇒ 00:15:18.700 Casie Aviles: Wanted to do.
147 00:15:18.700 ⇒ 00:15:20.620 Miguel de Veyra: So we have, yeah.
148 00:15:21.050 ⇒ 00:15:24.619 Casie Aviles: Yeah, we wanted to chat over the meetings right?
149 00:15:24.620 ⇒ 00:15:27.749 Miguel de Veyra: Yeah, okay, so let’s just say, let’s just do
150 00:15:28.980 ⇒ 00:15:32.459 Miguel de Veyra: like small scale. Let’s just do the meetings for ABC,
151 00:15:33.820 ⇒ 00:15:41.659 Miguel de Veyra: Yeah, let’s put that into basically, oh, right Zoom Meetings.
152 00:15:42.480 ⇒ 00:15:46.090 Miguel de Veyra: So wait. Let me share my screen. So we will need
153 00:15:46.800 ⇒ 00:15:48.560 Miguel de Veyra: you will need to create like a
154 00:15:49.370 ⇒ 00:15:52.629 Miguel de Veyra: a database, a table. And can you guys see my screen? Sorry
155 00:15:52.630 ⇒ 00:15:53.730 Amber Lin: Yeah, I can
156 00:15:54.360 ⇒ 00:16:00.159 Miguel de Veyra: So we need to create this. I think we do have some sort of this already. Right, Casey
157 00:16:01.348 ⇒ 00:16:03.790 Casie Aviles: Yeah, actually, I did implement
158 00:16:04.540 ⇒ 00:16:10.670 Casie Aviles: a vector store already a while ago, when I was initially working on the Zoom summarizer workflow
159 00:16:11.070 ⇒ 00:16:15.010 Miguel de Veyra: Yeah, okay, nice. So basically, the fields are this right?
160 00:16:15.420 ⇒ 00:16:16.260 Miguel de Veyra: The client
161 00:16:16.260 ⇒ 00:16:21.169 Casie Aviles: Yeah, the fields are not matching this one, so I guess we have to modify it
162 00:16:21.700 ⇒ 00:16:28.230 Miguel de Veyra: Yep, and then the in the automation part it. It doesn’t have to be anything, I think, I added, here.
163 00:16:28.400 ⇒ 00:16:34.400 Miguel de Veyra: specify an internal scheduling, or via any, and via
164 00:16:35.700 ⇒ 00:16:38.009 Miguel de Veyra: Windmill. This is what you’re gonna use right
165 00:16:39.030 ⇒ 00:16:43.020 Casie Aviles: Yeah, I use for a majority of the automation
166 00:16:44.420 ⇒ 00:16:51.030 Miguel de Veyra: Okay, okay, so basically, what’s gonna happen is, once it comes in, it has to go through here. First.st
167 00:16:53.920 ⇒ 00:16:59.580 Miguel de Veyra: Yeah. So we’re gonna process it. We’re gonna save the stuff here. And then we’re gonna process the store, the
168 00:17:00.130 ⇒ 00:17:04.199 Miguel de Veyra: the video file in zoom and then store it here. And then we’re gonna store the
169 00:17:04.200 ⇒ 00:17:04.780 Amber Lin: Hmm.
170 00:17:04.780 ⇒ 00:17:05.770 Miguel de Veyra: And Sd.
171 00:17:06.480 ⇒ 00:17:07.900 Miguel de Veyra: So now, if the boss
172 00:17:07.900 ⇒ 00:17:08.700 Amber Lin: Seat
173 00:17:08.700 ⇒ 00:17:15.979 Miguel de Veyra: Yeah. So now, if the bot asks, for example, so we’re gonna create a ui for this. So if the bot asks, you know.
174 00:17:18.220 ⇒ 00:17:26.630 Miguel de Veyra: So if the bot let’s say, asks, Hey, what’s the you know? Can you pull up this meeting ideally? What could happen
175 00:17:26.770 ⇒ 00:17:31.210 Miguel de Veyra: is that we also get like the video link, and they can even watch it in the Ui
176 00:17:31.600 ⇒ 00:17:33.330 Amber Lin: Oh!
177 00:17:34.250 ⇒ 00:17:36.470 Miguel de Veyra: So wait! Let me.
178 00:17:36.990 ⇒ 00:17:38.540 Miguel de Veyra: What the hell
179 00:17:39.370 ⇒ 00:17:43.210 Amber Lin: So web. Sorry. Windmill is kind of like Nan
180 00:17:46.590 ⇒ 00:17:48.840 Casie Aviles: It’s like orchestration.
181 00:17:48.840 ⇒ 00:17:51.399 Miguel de Veyra: Yeah, it’s more of like a hosting thing
182 00:17:51.400 ⇒ 00:17:54.359 Casie Aviles: Yeah, you could. I could create like 5
183 00:17:54.360 ⇒ 00:17:57.310 Amber Lin: What does that mean? I see. Okay.
184 00:17:58.690 ⇒ 00:18:00.619 Miguel de Veyra: Don’t worry, Amber. I don’t use windmill
185 00:18:02.017 ⇒ 00:18:17.499 Amber Lin: Cause I’m trying to. I’m trying to write out like a AI. Q. Fmq. Like, frequently asked questions for our sales team, and I like going through the weeds like, Oh, shit! What does this even mean? There’s so many applications
186 00:18:17.770 ⇒ 00:18:23.989 Miguel de Veyra: Sure create new. Wait. Let’s just say, internal. Q. 2. Internal.
187 00:18:24.210 ⇒ 00:18:26.140 Miguel de Veyra: Q. 2, q. 2.
188 00:18:26.720 ⇒ 00:18:28.010 Miguel de Veyra: We are internal.
189 00:18:29.310 ⇒ 00:18:34.219 Miguel de Veyra: So the way it will work is, basically, I love sticking notes, but sorry.
190 00:18:34.440 ⇒ 00:18:35.720 Miguel de Veyra: So
191 00:18:36.080 ⇒ 00:18:36.324 Amber Lin: Hmm.
192 00:18:36.570 ⇒ 00:18:38.410 Miguel de Veyra: Hey? Utah so trigger
193 00:18:42.700 ⇒ 00:18:43.630 Uttam Kumaran: Hey!
194 00:18:44.780 ⇒ 00:18:45.324 Amber Lin: Hello!
195 00:18:46.250 ⇒ 00:18:47.270 Miguel de Veyra: Hello, autumn.
196 00:18:54.420 ⇒ 00:19:02.869 Amber Lin: Oh, S. 3 for big file super for vectors.
197 00:19:03.690 ⇒ 00:19:04.770 Miguel de Veyra: And this is not
198 00:19:05.620 ⇒ 00:19:06.890 Amber Lin: Snowflake.
199 00:19:06.890 ⇒ 00:19:09.450 Miguel de Veyra: So yeah, it’s basically gonna look something like this.
200 00:19:09.610 ⇒ 00:19:12.520 Miguel de Veyra: So windmill store, the file uploading
201 00:19:13.280 ⇒ 00:19:15.829 Amber Lin: Wait? Where does wind? Where does windmill go
202 00:19:15.830 ⇒ 00:19:16.610 Miguel de Veyra: Trigger.
203 00:19:17.600 ⇒ 00:19:19.790 Amber Lin: Oh, can we write in there? Please.
204 00:19:26.580 ⇒ 00:19:32.540 Miguel de Veyra: I talked to them. Lade Putham Super is left, or
205 00:19:34.080 ⇒ 00:19:35.150 Uttam Kumaran: Okay. Cool.
206 00:19:35.840 ⇒ 00:19:40.389 Miguel de Veyra: And then snow is the dash afterwards. Okay.
207 00:19:40.800 ⇒ 00:19:45.820 Amber Lin: Oh, okay, alright. And then where does the ui go in?
208 00:19:46.122 ⇒ 00:19:49.750 Miguel de Veyra: That’s for later. That’s for later. Now, we’ll just do this. Yeah.
209 00:19:50.220 ⇒ 00:20:01.770 Amber Lin: Okay, yeah. Well, Miguel made very, very detailed tickets for the Zoom and the slack one. And then we talked about. And I and I asked, okay, so what are we going to deliver this week? And I think
210 00:20:01.920 ⇒ 00:20:06.540 Amber Lin: we’re gonna have a small scale for ABC. Is that right, Miguel?
211 00:20:06.880 ⇒ 00:20:14.150 Miguel de Veyra: Yeah. So the idea is like, for example, we wanna basically create by the end of this week, just the Zoom Meetings
212 00:20:14.370 ⇒ 00:20:20.550 Miguel de Veyra: for ABC, you know, and then be able to chat with that with that, does that work for you?
213 00:20:21.870 ⇒ 00:20:23.999 Uttam Kumaran: Yeah, I would probably pick
214 00:20:24.290 ⇒ 00:20:26.200 Miguel de Veyra: Or you know any any client
215 00:20:26.430 ⇒ 00:20:34.979 Uttam Kumaran: Yeah, I guess I would pick. I would pick one of the data clients because they’re like very much in need for some of this basic questions
216 00:20:34.980 ⇒ 00:20:35.639 Miguel de Veyra: Okay. Okay.
217 00:20:35.970 ⇒ 00:20:40.770 Uttam Kumaran: So I would just deliver for them. Because you guys are, you’ll be able to deliver the ABC stuff pretty easily.
218 00:20:41.250 ⇒ 00:20:48.523 Amber Lin: Yeah, maybe for Eden or Javi, whichever ones in a dire situation
219 00:20:49.280 ⇒ 00:20:53.220 Uttam Kumaran: Yeah, you can. Again, I would ask Demolati, or wish to pick one
220 00:20:53.220 ⇒ 00:20:53.570 Amber Lin: Okay.
221 00:20:53.570 ⇒ 00:20:55.459 Miguel de Veyra: Okay, yeah. I’ll just ask them loudly. Then
222 00:20:55.770 ⇒ 00:20:56.440 Uttam Kumaran: Okay.
223 00:20:57.120 ⇒ 00:21:05.290 Miguel de Veyra: And then, yeah, that is pretty much it. Of course, this one. Ideally, we wanna finish this by Wednesday or even
224 00:21:05.500 ⇒ 00:21:06.749 Miguel de Veyra: I’m not sure if it’s true.
225 00:21:07.330 ⇒ 00:21:10.139 Miguel de Veyra: finishable by Wednesday. Casey, what do you think? How many
226 00:21:10.140 ⇒ 00:21:10.880 Amber Lin: I mean
227 00:21:10.880 ⇒ 00:21:11.790 Miguel de Veyra: Yeah. Estimate?
228 00:21:11.790 ⇒ 00:21:15.590 Amber Lin: Aren’t you guys on holiday today and tomorrow
229 00:21:15.590 ⇒ 00:21:17.189 Miguel de Veyra: No, no, only today, supposedly.
230 00:21:17.190 ⇒ 00:21:19.019 Amber Lin: Okay, so, okay.
231 00:21:20.510 ⇒ 00:21:22.880 Miguel de Veyra: How many points do you think this would take, Casey
232 00:21:23.510 ⇒ 00:21:26.370 Casie Aviles: How do we define points? I’m sorry. I’m not sure
233 00:21:26.990 ⇒ 00:21:28.869 Miguel de Veyra: Oh, wait! I think Utam sent it to me.
234 00:21:29.200 ⇒ 00:21:30.700 Miguel de Veyra: Amber, do you have that on hand?
235 00:21:30.700 ⇒ 00:21:32.459 Uttam Kumaran: Yeah, there’s a it’s in the. It’s in no
236 00:21:32.460 ⇒ 00:21:34.670 Amber Lin: I’m gonna pull it up. Yes.
237 00:21:34.950 ⇒ 00:21:50.650 Amber Lin: I will read it to you. So 1 point is an hour very simple. No one knows. 2 points is 2 to 3 h, 3 points is 4 to 5 h, and 5 points is 6 to 8 HA full day of work
238 00:21:51.130 ⇒ 00:21:52.910 Miguel de Veyra: 8 is on it. 8 is what
239 00:21:53.150 ⇒ 00:21:55.769 Amber Lin: It is 1.5 to 2 days
240 00:21:56.000 ⇒ 00:21:57.630 Miguel de Veyra: 8.th I think this is 8
241 00:21:59.660 ⇒ 00:22:00.620 Amber Lin: Number, 13,
242 00:22:00.620 ⇒ 00:22:03.750 Uttam Kumaran: So I would rather us break. I would rather us break it up.
243 00:22:06.770 ⇒ 00:22:09.319 Uttam Kumaran: Okay, so this should be like sub issues. Then
244 00:22:10.600 ⇒ 00:22:16.620 Uttam Kumaran: either sub issues, or you break it up into to separate tickets, overall
245 00:22:17.430 ⇒ 00:22:18.390 Miguel de Veyra: Okay. Okay.
246 00:22:20.230 ⇒ 00:22:21.090 Uttam Kumaran: Because.
247 00:22:22.240 ⇒ 00:22:28.889 Uttam Kumaran: yeah, otherwise, it’s like we, we wanna basically have things closer to like 5 points or less for the most part
248 00:22:28.890 ⇒ 00:22:33.649 Miguel de Veyra: Okay, okay, yeah, that makes sense. So it’s like, it’s not too broad.
249 00:22:35.720 ⇒ 00:22:37.759 Miguel de Veyra: Okay, yeah, let me think about that.
250 00:23:05.600 ⇒ 00:23:08.070 Miguel de Veyra: So I would say, these are like the 4 tickets
251 00:23:11.550 ⇒ 00:23:15.009 Miguel de Veyra: sub tickets. I think this is a good time to use sub gracious here.
252 00:23:18.100 ⇒ 00:23:21.649 Miguel de Veyra: So yeah, I think of, we can start with this
253 00:23:27.120 ⇒ 00:23:29.329 Miguel de Veyra: I’ll add the tickets after this meeting.
254 00:23:31.850 ⇒ 00:23:36.630 Miguel de Veyra: and then uttan, by the way, in terms of like the slack messages. Oh.
255 00:23:37.480 ⇒ 00:23:44.350 Miguel de Veyra: as discussed, I don’t think we need to store the messages in S. 3 for this. We could just store them in
256 00:23:44.750 ⇒ 00:23:47.150 Miguel de Veyra: super base, like the textual stuff.
257 00:23:48.720 ⇒ 00:23:50.309 Miguel de Veyra: Is that something you agree with
258 00:24:05.630 ⇒ 00:24:07.330 Amber Lin: Youtube’s on mute
259 00:24:09.490 ⇒ 00:24:12.748 Uttam Kumaran: Oh, sorry, I I guess what I what I was saying is
260 00:24:13.270 ⇒ 00:24:18.600 Uttam Kumaran: You can consider superbase and snowflake as use case based applications
261 00:24:18.600 ⇒ 00:24:19.000 Miguel de Veyra: Yep.
262 00:24:19.000 ⇒ 00:24:22.579 Uttam Kumaran: Like use case based databases. Right? So snowflake is for analytics
263 00:24:22.580 ⇒ 00:24:23.000 Miguel de Veyra: Yes.
264 00:24:23.000 ⇒ 00:24:40.930 Uttam Kumaran: And Supa base is for vector store. But we need to have a central place where all this data lands first.st So everything has to get put into S. 3, because we may replace super base with something. Tomorrow we may replace Snowflake with something tomorrow, so those can’t be like the source of truth.
265 00:24:41.080 ⇒ 00:25:04.040 Uttam Kumaran: You basically want to create what’s called a data lake which is unstructured storage or semi structured storage of all of our data. So you can basically just have folders and some sort of sub folder structure for each core data source, whether it’s zoom and then like the meeting name. And it’s just a list of those, but that’s the best architecture right now is just to dump it all. There
266 00:25:04.040 ⇒ 00:25:06.800 Miguel de Veyra: Okay? Okay? Oh, yeah. And then, in terms of like.
267 00:25:07.290 ⇒ 00:25:10.569 Miguel de Veyra: in terms of like structuring it.
268 00:25:10.978 ⇒ 00:25:19.629 Miguel de Veyra: Is it? Do you want it to be, I mean, we just I discussed it with them a lot, and we think you know client, for example, ABC should have
269 00:25:20.020 ⇒ 00:25:31.449 Miguel de Veyra: sub folders a fold, a sub folder, called ABC should have their own slack messages, emails. Or should we stick to our current plan where where it everything is? Basically there
270 00:25:32.610 ⇒ 00:25:46.219 Uttam Kumaran: I mean, it depends like, this is like a. This is more of like an architecture. Decision is that, are you gonna dump everything 1st and then reorganize it like, it depends on. If you’re gonna classify stuff before S. 3,
271 00:25:46.500 ⇒ 00:25:53.440 Uttam Kumaran: right? Like, if you’re not, if you’re basically not going to classify stuff until after S. 3. Then
272 00:25:53.680 ⇒ 00:25:55.870 Uttam Kumaran: you just need it all dumped somewhere.
273 00:25:56.810 ⇒ 00:26:00.759 Uttam Kumaran: like, right? You don’t need to classify it first, st
274 00:26:00.970 ⇒ 00:26:01.540 Miguel de Veyra: Yeah.
275 00:26:01.960 ⇒ 00:26:08.020 Uttam Kumaran: You can classify it. Second. So that’s probably what I would suggest is, don’t do anything to the data
276 00:26:08.250 ⇒ 00:26:09.080 Uttam Kumaran: until after
277 00:26:09.332 ⇒ 00:26:10.090 Miguel de Veyra: But that’s great.
278 00:26:10.410 ⇒ 00:26:12.880 Miguel de Veyra: Okay, yeah, that makes sense. That’s that guys
279 00:26:13.280 ⇒ 00:26:17.380 Uttam Kumaran: You want. You want to basically like a raw source of truth for everything
280 00:26:17.970 ⇒ 00:26:20.470 Miguel de Veyra: Before you make any modifications.
281 00:26:23.760 ⇒ 00:26:27.570 Miguel de Veyra: Okay, yeah, that makes a lot of sense. We’ll, I’ll change this up.
282 00:26:29.996 ⇒ 00:26:31.689 Miguel de Veyra: Da, da.
283 00:26:39.390 ⇒ 00:26:46.260 Miguel de Veyra: okay, yeah. I think I think that is pretty clear to me. I’ll I’ll just work on this before
284 00:26:46.840 ⇒ 00:26:55.139 Miguel de Veyra: I go out and finalize this, and then I’ll tag you them to review it, and then hopefully, we can move this to ready for development by end of day.
285 00:26:55.140 ⇒ 00:27:00.239 Uttam Kumaran: Okay, yeah. And then and then, can you explain this the way we’re doing slack messages
286 00:27:02.150 ⇒ 00:27:03.859 Uttam Kumaran: or the what the proposed way
287 00:27:04.772 ⇒ 00:27:12.009 Miguel de Veyra: Yeah. So basically it would be very straightforward. We basically just, you know, a trigger and windmill.
288 00:27:12.420 ⇒ 00:27:14.459 Miguel de Veyra: We’re gonna use windmill here, Casey, right?
289 00:27:15.450 ⇒ 00:27:18.800 Uttam Kumaran: My my question is, why, why do this via webhook?
290 00:27:19.360 ⇒ 00:27:22.380 Uttam Kumaran: If I was to take the other side like, why do this via web hook
291 00:27:25.450 ⇒ 00:27:30.500 Uttam Kumaran: like. Why not just pull the Api for all the messages
292 00:27:32.550 ⇒ 00:27:35.590 Miguel de Veyra: Oh, yeah. Cause we also need the existing messages. Right?
293 00:27:36.990 ⇒ 00:27:41.780 Uttam Kumaran: Like think about, I guess. Think about like, what’s the use case for batch versus?
294 00:27:43.720 ⇒ 00:27:49.980 Uttam Kumaran: For batch versus web hook here, like, if we need the data real time
295 00:27:50.770 ⇒ 00:27:52.579 Uttam Kumaran: Then webhook makes sense.
296 00:27:52.740 ⇒ 00:27:53.900 Uttam Kumaran: Otherwise
297 00:27:54.490 ⇒ 00:27:59.480 Uttam Kumaran: you should just bash it because we’re gonna pay for web. But we’re going to pay for windmill to be up, 24, 7,
298 00:27:59.980 ⇒ 00:28:06.739 Miguel de Veyra: Okay, okay, like, basically like the one we do in the leadership channel, where every day
299 00:28:07.000 ⇒ 00:28:08.269 Miguel de Veyra: it gets all the messages
300 00:28:08.270 ⇒ 00:28:13.269 Uttam Kumaran: Yeah, I mean, it’s up to you to think about what is the requirements from the client on the timing. But
301 00:28:14.030 ⇒ 00:28:16.990 Uttam Kumaran: I don’t feel like we have a use case.
302 00:28:19.990 ⇒ 00:28:23.482 Uttam Kumaran: Sorry I don’t feel like we have a use case. That’s
303 00:28:24.200 ⇒ 00:28:26.300 Uttam Kumaran: That needs real time.
304 00:28:26.300 ⇒ 00:28:27.400 Miguel de Veyra: Okay. Yeah.
305 00:28:30.900 ⇒ 00:28:31.799 Uttam Kumaran: Okay, yeah.
306 00:28:31.800 ⇒ 00:28:38.460 Uttam Kumaran: again, this is what I would. This is what I would like. I would. I would ask them a lot. But basically, my guess is that
307 00:28:38.920 ⇒ 00:28:50.879 Uttam Kumaran: people are going to require at least a good place to start is like you have 24 h meaning anything that’s happened in the last 24 h is available, and then you can work on making that shorter and shorter. But I wouldn’t start with Web Hook, because
308 00:28:51.250 ⇒ 00:28:55.219 Uttam Kumaran: we’re gonna have to pay for windmill to be up basically all day. So
309 00:28:55.220 ⇒ 00:29:03.199 Miguel de Veyra: Okay, yeah, I mean, I was actually thinking about that because we already have that set up. But basically collects all the
310 00:29:03.430 ⇒ 00:29:08.099 Miguel de Veyra: messages for the day, and just dumps it into, you know. Hey? There’s no messages here
311 00:29:09.640 ⇒ 00:29:10.500 Casie Aviles: Schedule.
312 00:29:10.500 ⇒ 00:29:14.540 Miguel de Veyra: You know. Yeah, if I was like, huh, but if
313 00:29:14.540 ⇒ 00:29:19.810 Uttam Kumaran: But that’s why you so there, there’s gonna be 2 pieces, this. So one you have to backfill everything
314 00:29:20.060 ⇒ 00:29:21.460 Miguel de Veyra: Right, so
315 00:29:21.620 ⇒ 00:29:27.010 Uttam Kumaran: All, all messages ever sent, and again
316 00:29:29.690 ⇒ 00:29:33.680 Miguel de Veyra: He got cut off with him, but
317 00:29:33.680 ⇒ 00:29:38.509 Uttam Kumaran: Take the I don’t know. I would probably just start with
318 00:29:38.510 ⇒ 00:29:39.100 Miguel de Veyra: Gotcha
319 00:29:39.100 ⇒ 00:29:45.650 Uttam Kumaran: Client channel, or like the relevant channels related to the client, and start there.
320 00:29:47.740 ⇒ 00:29:48.300 Miguel de Veyra: Okay.
321 00:29:48.870 ⇒ 00:29:56.909 Uttam Kumaran: I would probably just start there and then, and then what you can do is bring that, bring all that data in, and then make sure that it’s starting to get updated on a batch.
322 00:29:58.380 ⇒ 00:29:58.830 Miguel de Veyra: Okay.
323 00:29:58.830 ⇒ 00:29:59.510 Uttam Kumaran: You know.
324 00:30:02.270 ⇒ 00:30:10.900 Uttam Kumaran: And this is this will be helpful to talk to a waste today, too. So I think today, if you guys can even have these ready. You can use the meeting with the data team to review.
325 00:30:11.050 ⇒ 00:30:14.259 Uttam Kumaran: like what the plan is because they’ll give you some.
326 00:30:14.440 ⇒ 00:30:17.100 Uttam Kumaran: Oashi is a data engineer. He’ll give you some like
327 00:30:17.550 ⇒ 00:30:20.330 Uttam Kumaran: insight into what the best way to run this
328 00:30:20.840 ⇒ 00:30:23.840 Miguel de Veyra: I think a wish is out for there. I’ll probably ask them, Ladi
329 00:30:24.510 ⇒ 00:30:25.120 Uttam Kumaran: Okay.
330 00:30:26.070 ⇒ 00:30:28.220 Miguel de Veyra: But, yeah, okay, yeah, very insightful.
331 00:30:29.260 ⇒ 00:30:30.490 Miguel de Veyra: I’ll update this
332 00:30:31.950 ⇒ 00:30:40.059 Amber Lin: Yeah, and we’re having the meeting with the data team later to go. So if you have any questions you want to ask from real time, we can
333 00:30:40.710 ⇒ 00:30:41.350 Miguel de Veyra: Okay. Yeah.
334 00:30:41.350 ⇒ 00:30:42.209 Amber Lin: I’m just, yeah.
335 00:30:42.568 ⇒ 00:30:45.789 Miguel de Veyra: Is, gonna be there. But let me check actually
336 00:30:46.160 ⇒ 00:30:47.330 Amber Lin: Like both of them.
337 00:30:47.870 ⇒ 00:30:48.409 Miguel de Veyra: Okay.
338 00:30:49.300 ⇒ 00:30:51.839 Miguel de Veyra: Should we bring Annie into this meeting, too, or no?
339 00:30:55.520 ⇒ 00:30:58.350 Miguel de Veyra: Cause she’s gonna be working on ABC stuff right?
340 00:30:58.510 ⇒ 00:30:59.060 Miguel de Veyra: Or nurse
341 00:30:59.060 ⇒ 00:31:02.639 Uttam Kumaran: Yeah, I mean, it’s it’s up to you guys. I I would say, like.
342 00:31:02.830 ⇒ 00:31:08.600 Uttam Kumaran: probably like, it’s probably best to to treat Annie as like someone who’s gonna
343 00:31:08.600 ⇒ 00:31:10.510 Amber Lin: Using the tool
344 00:31:10.510 ⇒ 00:31:13.639 Uttam Kumaran: But not but not like the core product owner.
345 00:31:13.640 ⇒ 00:31:14.130 Miguel de Veyra: Okay. Great.
346 00:31:14.130 ⇒ 00:31:18.330 Uttam Kumaran: Because Demolata, in a way. Ultimately I’ll ask them to approve whether it’s
347 00:31:18.520 ⇒ 00:31:30.449 Uttam Kumaran: it’s been like done to their for their needs. Right? So I think they’re the number one priority to make sure that they can use it, because they’ll also own scaling this up to the rest of the team.
348 00:31:33.740 ⇒ 00:31:35.149 Miguel de Veyra: Okay, not at the least.
349 00:31:35.780 ⇒ 00:31:40.449 Uttam Kumaran: Okay, what are the other like complicated sources that we have? So we have zoom.
350 00:31:40.620 ⇒ 00:31:48.280 Uttam Kumaran: we have. So the for for the zoom, you’re gonna just drop all the video audio text files into there first, st
351 00:31:48.280 ⇒ 00:31:48.970 Miguel de Veyra: Yeah.
352 00:31:50.650 ⇒ 00:31:51.290 Uttam Kumaran: Okay.
353 00:31:51.540 ⇒ 00:31:57.779 Miguel de Veyra: And then I think I would say, because the ones here are Github emails and linear tickets
354 00:31:58.490 ⇒ 00:32:05.709 Uttam Kumaran: Yeah, I would say, like, Github is probably next in terms of like, how how helpful it’s gonna be for the bot.
355 00:32:06.410 ⇒ 00:32:14.449 Uttam Kumaran: It’s gonna be Github. Next emails is like the lowest slack, is probably the highest
356 00:32:14.570 ⇒ 00:32:16.449 Uttam Kumaran: slack and Zoom are the highest
357 00:32:17.470 ⇒ 00:32:23.660 Miguel de Veyra: So, okay, slack zoom, Github, linear. And then emails, okay, yeah.
358 00:32:24.390 ⇒ 00:32:25.140 Uttam Kumaran: Yeah.
359 00:32:25.360 ⇒ 00:32:26.630 Miguel de Veyra: But again like
360 00:32:27.090 ⇒ 00:32:33.469 Uttam Kumaran: It’s like, you guys, could, you guys may be able to get pretty far by just building the agent purely off of
361 00:32:33.610 ⇒ 00:32:39.580 Uttam Kumaran: slack and zoom, because we already we already are doing. We basically have a lot of that already.
362 00:32:39.960 ⇒ 00:32:40.420 Amber Lin: Oh!
363 00:32:40.713 ⇒ 00:32:44.529 Uttam Kumaran: So I don’t know. I think that could be pretty interesting to see
364 00:32:44.930 ⇒ 00:32:45.350 Miguel de Veyra: Yeah, yeah.
365 00:32:45.350 ⇒ 00:32:52.089 Uttam Kumaran: Basically, you get it to the point where it’s just built on that. And then you, you can test whether we even need the other stuff, or we can deprioritize it right
366 00:32:52.290 ⇒ 00:32:52.650 Miguel de Veyra: Yes.
367 00:32:52.650 ⇒ 00:32:55.720 Uttam Kumaran: This is the game. This is sort of the game right now, because it’s like
368 00:32:55.870 ⇒ 00:33:01.000 Uttam Kumaran: we could spend another 3 weeks load everything in. But then we find that 2 of the sources are really like
369 00:33:01.380 ⇒ 00:33:04.589 Uttam Kumaran: the only one. The only alpha is in 2 of the sources, you know.
370 00:33:04.590 ⇒ 00:33:06.999 Miguel de Veyra: Yeah, like, 95% is there
371 00:33:07.380 ⇒ 00:33:08.290 Uttam Kumaran: Yeah.
372 00:33:09.220 ⇒ 00:33:10.500 Miguel de Veyra: Okay. Yeah. Sure.
373 00:33:13.900 ⇒ 00:33:15.430 Uttam Kumaran: Okay? What else?
374 00:33:15.430 ⇒ 00:33:18.030 Miguel de Veyra: So. I think.
375 00:33:22.040 ⇒ 00:33:24.579 Miguel de Veyra: Amber. I think that’s all for me right now.
376 00:33:30.730 ⇒ 00:33:39.060 Amber Lin: okay, I think this week we’ll have some work on ABC, mostly.
377 00:33:39.771 ⇒ 00:34:05.449 Amber Lin: adjusting the bot performance based on the feedback, I’m getting the rollout going. So that’s going pretty well. I’ll need. Maybe Casey’s a few hours of Casey’s time to adjust the bot performance, and then I’ll work with Annie. Maybe Casey Annie will have to work together to get the Evals and error ratings correct, because then it’s not really reflecting our performance right now. And I think that’s a big.
378 00:34:05.760 ⇒ 00:34:18.140 Amber Lin: that’s a big blocker. I want that to be accurate. So the client can know that we’re actually doing pretty well, because right now the error rates are not. They’re they’re pretty high. And the
379 00:34:18.639 ⇒ 00:34:25.319 Amber Lin: what is it? The quality score rating is also probably not reflective of how great we’re doing so.
380 00:34:26.020 ⇒ 00:34:26.320 Amber Lin: Yes.
381 00:34:26.320 ⇒ 00:34:38.851 Uttam Kumaran: Probably a good discussion, too, with on that data meeting today, like loop. Those like those data guys are are good. You could loop them into what the problem is. It’ll build some empathy for, like what we’re trying to do and
382 00:34:39.159 ⇒ 00:34:40.099 Amber Lin: Yeah, okay.
383 00:34:40.100 ⇒ 00:34:43.260 Uttam Kumaran: Yeah, ideally, Annie can sort of be in charge of like
384 00:34:43.510 ⇒ 00:34:52.710 Uttam Kumaran: Annie can sort of be in charge of like just taking a look at the data and then basically handing it over to AI team to say, Hey, this, this needs to be. This needs to be corrected.
385 00:34:54.409 ⇒ 00:34:59.603 Uttam Kumaran: That way. It kind of creates some separation. And then again, it’s like something that
386 00:35:00.139 ⇒ 00:35:06.619 Uttam Kumaran: you can take it out and then have as part of the sprint, basically like, Hey, we’re working on fixing up this Eval data.
387 00:35:07.031 ⇒ 00:35:14.309 Uttam Kumaran: But is she? She’s able to get into real and and see everything. I know she’s having some issues with Snowflake that I’ll resolve today
388 00:35:17.320 ⇒ 00:35:19.419 Amber Lin: Is she able to get into everything else?
389 00:35:23.060 ⇒ 00:35:26.210 Amber Lin: No other messages based on that
390 00:35:26.210 ⇒ 00:35:26.570 Uttam Kumaran: Okay.
391 00:35:26.866 ⇒ 00:35:28.350 Amber Lin: So I assume it’s good
392 00:35:29.430 ⇒ 00:35:30.040 Uttam Kumaran: Okay.
393 00:35:30.620 ⇒ 00:35:31.330 Amber Lin: Yeah.
394 00:35:36.530 ⇒ 00:35:41.409 Amber Lin: yeah, great. I think that’s all from my side, too.
395 00:35:41.910 ⇒ 00:35:43.550 Amber Lin: So probably
396 00:35:43.890 ⇒ 00:35:57.789 Amber Lin: maybe after I I think Wednesday will be pretty busy with this. So maybe, but Wednesday I’ll probably add the tickets, and Annie will take a look at it, and then Thursday we’ll have some time to work on BC.
397 00:35:59.040 ⇒ 00:35:59.610 Uttam Kumaran: Okay.
398 00:36:00.030 ⇒ 00:36:02.219 Amber Lin: Yeah, great, that’s all.
399 00:36:02.220 ⇒ 00:36:09.410 Uttam Kumaran: Yeah. And then let’s any any discussion on data or tickets just tag me, or we can do it. We can also review in slack.
400 00:36:10.750 ⇒ 00:36:16.900 Uttam Kumaran: you know, especially as cause some of these data problems like this is what we do on the data side I would love for us to.
401 00:36:17.190 ⇒ 00:36:23.410 Uttam Kumaran: you know, just have the discussion in public with everybody in slack. Or I know you guys met. So that’s great, too, like
402 00:36:23.440 ⇒ 00:36:47.035 Uttam Kumaran: as much as we can rely on the data team for any data questions. It’ll solve the problem. And then again, our goal is one of the I think, to think about for the meeting later today with the data team. Part of it, again, is to just ask them, like, what are the common questions that people are asking? How can we use a like? How can they start to think about opportunities for automation in in their work?
403 00:36:47.440 ⇒ 00:36:51.029 Uttam Kumaran: you know, and they may not know the things that that are possible.
404 00:36:51.160 ⇒ 00:37:03.280 Uttam Kumaran: But one of the things we want to note down is at the end of this exercise with them. If we have a list of like. Okay, they gave us 15 or 20 questions that currently will take 30 min to an hour to to answer.
405 00:37:03.760 ⇒ 00:37:14.190 Uttam Kumaran: And using the AI agent. It’s down to a few seconds. Right? So the that’s like the success story. So if you work yourself, if you work your way backwards from the success story.
406 00:37:15.080 ⇒ 00:37:18.180 Amber Lin: It makes it easy to know where we want to go. Right, like, let’s say in a month.
407 00:37:18.960 ⇒ 00:37:29.569 Uttam Kumaran: You would say, Okay, cool this where we crushed it for their team. What are what’s like? What is the story that’s printed on like what we did for them. Okay, you know. And sort of thinking through
408 00:37:29.680 ⇒ 00:37:47.129 Uttam Kumaran: great like, we’re able to answer 95% of their basic questions in 5 seconds or less, we’re able to then answer 50% of their harder questions. Right? That’s the success story. So start there and think about what that story is, and that’ll give you a really good North Star for this
409 00:37:47.130 ⇒ 00:37:47.550 Amber Lin: Project.
410 00:37:47.550 ⇒ 00:37:48.560 Uttam Kumaran: With the data team.
411 00:37:49.073 ⇒ 00:37:59.080 Uttam Kumaran: You know, it’s really common like this is pretty similar to what Amazon does which they write. It’s called Pr. FAQ, which is basically before any project starts, they write the press release.
412 00:37:59.620 ⇒ 00:38:07.790 Uttam Kumaran: So they write. They they write like what is going to get published in the press when the project launches even for internal stuff, and then they write an FAQ,
413 00:38:08.100 ⇒ 00:38:29.800 Uttam Kumaran: right? There’s a lot to that, because and that’s what they circulate like. That’s what they send around. Do you know what the goal is before starting? Right? So commonly like for Miguel and Casey? Remember, when we started this project. We were just like sort of going, and I was sort of setting the roadmap, and we were going, and we ended up in a in a decent place. But we still didn’t hit the target.
414 00:38:29.910 ⇒ 00:38:32.060 Uttam Kumaran: So this one we have to do differently.
415 00:38:32.240 ⇒ 00:38:37.960 Uttam Kumaran: like we really have to focus on both. What is it that the end customer needs.
416 00:38:38.590 ⇒ 00:38:54.480 Uttam Kumaran: and making sure that we hit that target at the end of this exercise right last time, I think we built a lot of great infrastructure. I definitely, of course, the 3 of us see the vision, but it’s not up to the 3 of us, right? It’s up to the folks that are going to use the agent. So that’s what I would really think about is
417 00:38:54.660 ⇒ 00:39:00.910 Uttam Kumaran: making sure that at every step of the way you’re heading towards the the milestone that’s going to affect the customer. The most
418 00:39:06.760 ⇒ 00:39:07.480 Uttam Kumaran: cool
419 00:39:07.860 ⇒ 00:39:08.570 Miguel de Veyra: Yeah.
420 00:39:12.340 ⇒ 00:39:13.790 Uttam Kumaran: Okay, that’s all. I had
421 00:39:13.790 ⇒ 00:39:19.459 Miguel de Veyra: Okay, okay, will you be the? I know you won’t be on the data meeting later.
422 00:39:19.920 ⇒ 00:39:24.669 Uttam Kumaran: I’ll yeah, I’ll try. I just have a couple of other things to take care of. So
423 00:39:24.780 ⇒ 00:39:31.520 Uttam Kumaran: just mess. Just message me if it’s if it’s helpful for me to be there, or if it’s totally going off the rails, I’ll show
424 00:39:31.960 ⇒ 00:39:35.159 Miguel de Veyra: Yeah, yeah, I think it would be a good example. Test.
425 00:39:35.940 ⇒ 00:39:36.510 Uttam Kumaran: Okay.
426 00:39:36.510 ⇒ 00:39:37.760 Miguel de Veyra: How it’ll go without you
427 00:39:38.230 ⇒ 00:39:38.920 Uttam Kumaran: Yeah.
428 00:39:39.960 ⇒ 00:39:44.830 Uttam Kumaran: no, it’s good. I mean, I I again for me, like I want to play backup like I want to
429 00:39:45.580 ⇒ 00:39:50.139 Uttam Kumaran: do what I’m doing here, which is like, okay, you guys got like, 80% of the way there, let me just
430 00:39:50.400 ⇒ 00:39:54.559 Uttam Kumaran: give you like one or 2 more things. That’s a much better role for me to play
431 00:39:54.560 ⇒ 00:39:54.990 Miguel de Veyra: Yep.
432 00:39:54.990 ⇒ 00:40:02.490 Uttam Kumaran: Because otherwise I’m gonna miss the core. I’m gonna not gonna be able to help in that way. Right? So that’s where, like, I’m best utilized, but
433 00:40:02.610 ⇒ 00:40:07.919 Uttam Kumaran: otherwise those 2 will not be bought in like, if I’m in every meeting it, it’ll be me driving stuff, so
434 00:40:07.920 ⇒ 00:40:08.340 Miguel de Veyra: Yeah.
435 00:40:08.340 ⇒ 00:40:16.358 Uttam Kumaran: It’s even helpful for the meeting, and that you’re not gonna have me sort of like destroying the silence.
436 00:40:16.780 ⇒ 00:40:17.220 Uttam Kumaran: Someone
437 00:40:17.220 ⇒ 00:40:23.909 Uttam Kumaran: and someone will have. Someone will have to step up. Yeah. So that’s that’s a benefit. That’s not a curse.
438 00:40:23.910 ⇒ 00:40:28.640 Miguel de Veyra: Okay, okay, yeah. I’ll see you guys there later. Then
439 00:40:28.640 ⇒ 00:40:29.480 Uttam Kumaran: Okay. Okay.
440 00:40:30.117 ⇒ 00:40:33.330 Amber Lin: We’ll see you at the stack list. Stand up
441 00:40:33.480 ⇒ 00:40:35.140 Miguel de Veyra: Okay, everyone. Bye-bye.
442 00:40:35.140 ⇒ 00:40:35.500 Amber Lin: Okay.
443 00:40:35.500 ⇒ 00:40:35.900 Uttam Kumaran: Bye.
444 00:40:35.900 ⇒ 00:40:37.160 Amber Lin: Thank you. Guys. Bye.