Meeting Title: ABC Standup Date: 2025-06-23 Meeting participants: Awaish Kumar, Casie Aviles, Mustafa Raja, Amber Lin
WEBVTT
1 00:02:50.420 ⇒ 00:02:51.500 Mustafa Raja: Hey!
2 00:02:54.700 ⇒ 00:02:55.590 Casie Aviles: A.
3 00:02:57.970 ⇒ 00:02:58.759 Mustafa Raja: How are you?
4 00:02:59.960 ⇒ 00:03:01.560 Casie Aviles: Yeah. Doing. Good.
5 00:03:02.100 ⇒ 00:03:02.440 Mustafa Raja: Yeah.
6 00:03:15.410 ⇒ 00:03:16.470 Amber Lin: Hi! There!
7 00:03:16.790 ⇒ 00:03:17.230 Casie Aviles: Am.
8 00:03:17.230 ⇒ 00:03:18.090 Mustafa Raja: Hey!
9 00:03:19.710 ⇒ 00:03:21.859 Amber Lin: I’ll try to keep this one quick.
10 00:03:22.804 ⇒ 00:03:37.400 Amber Lin: I also invited away and Annie, but I don’t think their their tasks are more related on the data side. So if they join later, I can talk separately with them. Just quickly looking at our current cycle.
11 00:03:41.420 ⇒ 00:03:55.120 Amber Lin: yeah, Casey, I saw that you moved this one into the current cycle. I’ll I’ll help you add a few more requirements, and honestly, instead of filing all of them, I think we can focus on the inspector sheet.
12 00:03:55.250 ⇒ 00:03:58.610 Amber Lin: and I will add more requirements for that.
13 00:03:59.370 ⇒ 00:04:01.769 Casie Aviles: Oh, okay. Okay. Sure. Yeah.
14 00:04:02.200 ⇒ 00:04:07.670 Amber Lin: I was just gonna bring that up, but I forgot to make the ticket
15 00:04:14.920 ⇒ 00:04:19.050 Casie Aviles: At least for this this week. Right? The inspector sheet.
16 00:04:19.812 ⇒ 00:04:28.559 Amber Lin: Yeah, I think that narrows the focus, there’ll be a less work to do so i’ll say.
17 00:04:29.480 ⇒ 00:04:35.250 Amber Lin: let me see, I’m, gonna say, it’s for.
18 00:04:37.980 ⇒ 00:04:39.520 Amber Lin: Yeah, I’ll move.
19 00:04:39.700 ⇒ 00:04:53.329 Amber Lin: I’ll add more details to that. I’ll move that to next cycle and then think for
20 00:04:53.630 ⇒ 00:04:56.960 Amber Lin: for Mustafa. I know you sent me another update.
21 00:04:57.280 ⇒ 00:04:59.329 Amber Lin: Let me quickly check.
22 00:04:59.980 ⇒ 00:05:01.339 Amber Lin: Want these.
23 00:05:09.330 ⇒ 00:05:14.170 Amber Lin: Yeah. Any updates on anything you want me to check? Are they both in these 2 tickets?
24 00:05:14.170 ⇒ 00:05:19.289 Mustafa Raja: Yeah. Yeah, I added more comments in the in progress tickets.
25 00:05:20.500 ⇒ 00:05:22.549 Amber Lin: Hmm, sounds good. Okay.
26 00:05:22.550 ⇒ 00:05:27.030 Mustafa Raja: Yeah. Yeah. I’d love if it if we can talk about these. Now.
27 00:05:27.870 ⇒ 00:05:33.030 Amber Lin: Oh, sure! Here.
28 00:05:37.960 ⇒ 00:05:42.419 Amber Lin: Oh, I mean, if it doesn’t change anything, then I guess it’s fine.
29 00:05:42.710 ⇒ 00:05:44.720 Amber Lin: We can. We can start out by pulling.
30 00:05:44.720 ⇒ 00:05:47.999 Amber Lin: Yeah, dog and figure it out later. Sure. Okay.
31 00:05:48.660 ⇒ 00:05:50.160 Mustafa Raja: Yeah, yeah.
32 00:05:50.550 ⇒ 00:05:52.190 Amber Lin: Then for this.
33 00:06:05.250 ⇒ 00:06:10.310 Amber Lin: Okay, so it we will still send it in Google Chat. Right? So we’ll
34 00:06:10.990 ⇒ 00:06:15.010 Amber Lin: say something. It’s in progress, and I’ll send would send a new message when it’s done.
35 00:06:15.820 ⇒ 00:06:19.490 Mustafa Raja: Yeah, because we cannot update the message in Google chat.
36 00:06:19.490 ⇒ 00:06:20.020 Amber Lin: Hmm.
37 00:06:20.190 ⇒ 00:06:33.510 Mustafa Raja: To make it, you know more. Good. What we can do is we can delete our placeholder message and send a new message. That would be the response.
38 00:06:34.210 ⇒ 00:06:40.660 Amber Lin: Sure. I mean, we don’t even really have to delete the place placeholder message. We’ll just say it’s in progress, and we’ll send it. Send a new one.
39 00:06:40.660 ⇒ 00:06:41.100 Amber Lin: Wow!
40 00:06:41.100 ⇒ 00:06:41.780 Amber Lin: Done!
41 00:06:42.142 ⇒ 00:06:45.039 Mustafa Raja: Yeah, yeah, yeah, that’s that’s even better than.
42 00:06:45.350 ⇒ 00:06:46.040 Amber Lin: Okay.
43 00:06:46.290 ⇒ 00:06:47.940 Amber Lin: Awesome. Let.
44 00:06:47.940 ⇒ 00:06:48.430 Mustafa Raja: Yeah, okay.
45 00:06:49.365 ⇒ 00:06:49.930 Mustafa Raja: And.
46 00:06:51.649 ⇒ 00:06:58.470 Mustafa Raja: And for the this drag ticket that we have in through cycle and there’s nothing.
47 00:06:58.470 ⇒ 00:06:59.116 Amber Lin: Get there!
48 00:06:59.440 ⇒ 00:07:12.989 Mustafa Raja: Yeah, let me know if let me know if you want to implement this I talked to autumn about this and I he, autumn said that it’s your call. So currently it’s taking about
49 00:07:13.130 ⇒ 00:07:22.400 Mustafa Raja: 5 to 25 seconds. The AI agent to generate the response. So the
50 00:07:23.367 ⇒ 00:07:27.900 Mustafa Raja: purpose of this ticket would be to improve the response. Time.
51 00:07:30.480 ⇒ 00:07:31.799 Mustafa Raja: So let me know.
52 00:07:31.800 ⇒ 00:07:32.640 Amber Lin: We’re both.
53 00:07:32.640 ⇒ 00:07:33.570 Mustafa Raja: If we should.
54 00:07:34.360 ⇒ 00:07:37.889 Amber Lin: Is this for both the trainer Bot, and for Andy support.
55 00:07:37.890 ⇒ 00:07:40.370 Mustafa Raja: No, no, this is only for the trainer board.
56 00:07:46.830 ⇒ 00:07:48.560 Amber Lin: How long would this take.
57 00:07:52.865 ⇒ 00:07:56.190 Mustafa Raja: Half an hour, 45 min, I guess.
58 00:07:56.560 ⇒ 00:07:57.270 Amber Lin: Oh,
59 00:07:59.780 ⇒ 00:08:01.679 Amber Lin: We can try it out.
60 00:08:02.060 ⇒ 00:08:05.096 Casie Aviles: Can I ask if yeah. Sorry. Can I ask if
61 00:08:05.630 ⇒ 00:08:10.260 Casie Aviles: what exactly are we going to do? Rug on like?
62 00:08:10.540 ⇒ 00:08:16.840 Casie Aviles: That’s the knowledge? But I I mean, that’s a trainer, but require something to be vectorized. Yeah.
63 00:08:16.840 ⇒ 00:08:21.467 Mustafa Raja: Yeah, yes. So so what what currently is happening is
64 00:08:22.480 ⇒ 00:08:43.409 Mustafa Raja: how? How are we passing the context to the trainer? Bot, we are passing it by summarizing each section into one to 2 lines. So, and then and then what we do is we pull the whole thing all summaries of all sections. And then we put it in our trainer bot.
65 00:08:43.510 ⇒ 00:08:48.369 Mustafa Raja: So what happens is then it it becomes a very, very long message.
66 00:08:49.470 ⇒ 00:08:50.080 Casie Aviles: Oh!
67 00:08:50.080 ⇒ 00:08:53.579 Mustafa Raja: So what we want to do is so if we can reduce that
68 00:08:54.480 ⇒ 00:08:58.900 Mustafa Raja: then I feel I feel the the response. Time should be better.
69 00:08:59.460 ⇒ 00:09:06.170 Mustafa Raja: What this strategy is called is cag cache, augmented generation.
70 00:09:07.650 ⇒ 00:09:11.919 Casie Aviles: Okay, okay, yeah. So you’re, you’re just going to vectorize the central dock. Right?
71 00:09:12.500 ⇒ 00:09:22.029 Mustafa Raja: Yeah, yeah. But let me know. Let me know your opinion. If if this is, this is what we should do, because what we currently are doing
72 00:09:23.400 ⇒ 00:09:30.219 Mustafa Raja: what we currently are doing makes our trainer board very accurate in identifying the sections.
73 00:09:30.780 ⇒ 00:09:38.640 Mustafa Raja: So if we do, if we if we go with the rag the accuracy might get affected.
74 00:09:39.330 ⇒ 00:09:41.090 Casie Aviles: Okay, I see, yeah.
75 00:09:41.430 ⇒ 00:09:47.920 Mustafa Raja: Because you know it. It then depends on if it can retrieve the documents, the relevant documents or not.
76 00:09:49.460 ⇒ 00:09:49.840 Amber Lin: I see.
77 00:09:49.840 ⇒ 00:09:50.190 Casie Aviles: Okay.
78 00:09:50.190 ⇒ 00:09:52.480 Amber Lin: Are we doing rag for Andy.
79 00:09:53.160 ⇒ 00:09:56.546 Casie Aviles: No we’re doing. We’re just kind of
80 00:09:57.780 ⇒ 00:10:02.189 Casie Aviles: routing the contacts. Right? That’s what we’re doing right now.
81 00:10:02.590 ⇒ 00:10:03.260 Amber Lin: Oh!
82 00:10:03.260 ⇒ 00:10:10.700 Casie Aviles: Yes, so we’re taking, like the con, the the contents of the documents. And then we’re just, you know, kind of routing it. So it doesn’t
83 00:10:10.890 ⇒ 00:10:12.710 Casie Aviles: get everything but.
84 00:10:13.620 ⇒ 00:10:15.289 Casie Aviles: It’s still like, you know, just
85 00:10:15.630 ⇒ 00:10:18.869 Casie Aviles: putting it into context, but just with extra steps, I guess.
86 00:10:19.930 ⇒ 00:10:25.200 Mustafa Raja: Yeah, so so it’s pretty much the same. As the trainer button. Right?
87 00:10:25.310 ⇒ 00:10:30.679 Mustafa Raja: What? But what trained bot is getting is summaries of all sections of the document.
88 00:10:33.430 ⇒ 00:10:34.400 Casie Aviles: Okay, yeah.
89 00:10:34.400 ⇒ 00:10:41.079 Mustafa Raja: So so the so the main. So the main question for this track ticket would be, is it worth it?
90 00:10:49.950 ⇒ 00:10:55.970 Amber Lin: Let’s see, I think these 2 tickets are almost done right. How long would is left.
91 00:10:55.970 ⇒ 00:10:58.499 Mustafa Raja: The docs Api will will take some time.
92 00:10:59.505 ⇒ 00:11:06.589 Amber Lin: Okay, but will we time this week left after we finish those 2?
93 00:11:07.010 ⇒ 00:11:07.750 Amber Lin: Okay.
94 00:11:07.750 ⇒ 00:11:09.090 Mustafa Raja: Yeah.
95 00:11:10.100 ⇒ 00:11:14.929 Amber Lin: Yeah, I mean, if you said this one takes 45 min, I think it’s worth trying it out.
96 00:11:15.782 ⇒ 00:11:28.939 Amber Lin: If it’s a 45 min task. But I wouldn’t say it’s that high of a priority. We just need it to work now, and to see how we can compare it. And then well, if we get extra time we can do the rag.
97 00:11:30.080 ⇒ 00:11:30.920 Mustafa Raja: Okay.
98 00:11:31.280 ⇒ 00:11:31.800 Amber Lin: Yeah.
99 00:11:32.520 ⇒ 00:11:40.919 Mustafa Raja: Yeah. So the so so our goal would be to compare what we are, what our current strategy is with the rag. Then right.
100 00:11:41.500 ⇒ 00:11:46.130 Amber Lin: Yeah, exactly. So to see how much it improves and or like.
101 00:11:46.350 ⇒ 00:12:01.719 Amber Lin: for the amount of time that it improves, how much accuracy are we sacrificing? And can we tolerate that because we don’t know how how inaccurate it will become? But if it’s 45 min to test it out like sometime this week. We should. We can test that out.
102 00:12:02.870 ⇒ 00:12:05.199 Mustafa Raja: Yeah, yeah, that’s good.
103 00:12:09.620 ⇒ 00:12:10.310 Amber Lin: Okay.
104 00:12:14.810 ⇒ 00:12:27.480 Amber Lin: yeah. And Casey, I will flesh out that ticket. I realize that also has nothing in it. So I got the requirements from from Janice and Shannon and I. I will give you the details.
105 00:12:27.800 ⇒ 00:12:28.800 Casie Aviles: Okay. Sure.
106 00:12:28.800 ⇒ 00:12:30.689 Amber Lin: Okay. Thank you.
107 00:12:33.640 ⇒ 00:12:34.370 Casie Aviles: Alright. Thank you. Guys.
108 00:12:34.370 ⇒ 00:12:35.240 Amber Lin: Alright!
109 00:12:35.240 ⇒ 00:12:36.580 Mustafa Raja: Bye, bye. Thank you.