Meeting Title: ABC working session Date: 2026-02-17 Meeting participants: Mustafa Raja, Samuel Roberts, Amber Lin, Casie Aviles
WEBVTT
1 00:01:34.460 ⇒ 00:01:35.100 Samuel Roberts: Aye.
2 00:01:36.990 ⇒ 00:01:38.650 Mustafa Raja: Hey. Hey, how are you?
3 00:01:39.440 ⇒ 00:01:41.410 Samuel Roberts: Doing alright. How about yourself?
4 00:01:42.150 ⇒ 00:01:45.459 Mustafa Raja: Yeah, I’m just… I’ve been busy. This week is very busy.
5 00:01:46.040 ⇒ 00:01:46.730 Samuel Roberts: Yeah.
6 00:01:49.610 ⇒ 00:01:50.640 Samuel Roberts: Hamburg.
7 00:01:52.600 ⇒ 00:01:53.580 Amber Lin: Hello!
8 00:01:53.980 ⇒ 00:01:54.960 Mustafa Raja: Hey, how are you?
9 00:01:55.700 ⇒ 00:02:03.120 Amber Lin: I’m good. My internet is a little bit slow today, so I’ll try not to share a screen, but…
10 00:02:03.670 ⇒ 00:02:08.050 Amber Lin: I think… We should be good. Is Casey coming today?
11 00:02:08.240 ⇒ 00:02:10.430 Samuel Roberts: Okay, there we go. Looks like it.
12 00:02:10.440 ⇒ 00:02:11.200 Amber Lin: Okay.
13 00:02:11.590 ⇒ 00:02:23.230 Amber Lin: Okay. I know the zip code stuff’s going as planned, so probably don’t need to talk about it as much. The only comment is probably I’ll… I’ll see if…
14 00:02:23.560 ⇒ 00:02:38.159 Amber Lin: The CSRs is testing it, and if… if the feedback improved. But other than that, no comments on that side. So mostly, you want to talk about, Maestra, and then the central doc stuff.
15 00:02:39.580 ⇒ 00:02:40.150 Mustafa Raja: Yeah.
16 00:02:41.010 ⇒ 00:02:48.930 Mustafa Raja: So, for Central Doc stuff, I had a… I had, I had a pass over it, and Sam and I reviewed it.
17 00:02:51.270 ⇒ 00:03:10.840 Mustafa Raja: And we… we came up with a new… we wanted to try a new strategy. It’s mostly, it’s mostly just, we need to create a system smart enough that, it is always, or at least mostly, choosing the right sections to compare, right?
18 00:03:10.970 ⇒ 00:03:24.269 Mustafa Raja: Yeah, that’s pretty much it. So, the strategy now we want to take a look at is we are storing, section summaries in Superbase, so maybe we would want to use that and have an LLM
19 00:03:24.270 ⇒ 00:03:32.340 Mustafa Raja: look into, potential correct, pairs that LLM can then further compare.
20 00:03:32.490 ⇒ 00:03:37.699 Mustafa Raja: So… that’s an update on that, and I will be working on that today.
21 00:03:39.430 ⇒ 00:03:42.379 Mustafa Raja: I just have… I’ve been busy with, eating stuff.
22 00:03:42.660 ⇒ 00:03:44.019 Mustafa Raja: I have time today.
23 00:03:44.540 ⇒ 00:04:03.190 Amber Lin: Okay, I think before we dive in, because I think the central stuff might need some new planning, since you guys said there’s new tickets, I don’t really know if the plan we made before is still exactly the same. But before that, my question on the menstrual migration stuff is, okay.
24 00:04:03.190 ⇒ 00:04:07.580 Amber Lin: When are we able to get the clients to use it?
25 00:04:11.350 ⇒ 00:04:13.510 Samuel Roberts: Yeah, so it’s in a dev environment.
26 00:04:14.040 ⇒ 00:04:16.600 Samuel Roberts: I wanted to run some, like.
27 00:04:17.110 ⇒ 00:04:21.389 Samuel Roberts: Tests on that with the, like, live questions coming in first.
28 00:04:22.220 ⇒ 00:04:27.510 Samuel Roberts: If that goes well, which I hope it does, and I kind of expect it to, then we could probably…
29 00:04:28.400 ⇒ 00:04:36.529 Samuel Roberts: you know, depending on how many of those come through, if we can get that set up sooner rather than later, a few days of that, I think, would be fine to kind of confirm that it’s good.
30 00:04:37.110 ⇒ 00:04:37.830 Amber Lin: Okay.
31 00:04:37.830 ⇒ 00:04:40.880 Samuel Roberts: I don’t know. Mustafa Casey, how you feel about that?
32 00:04:41.060 ⇒ 00:04:45.799 Mustafa Raja: Yeah, I feel the same. I think, before rolling it out to the client.
33 00:04:45.940 ⇒ 00:04:50.809 Mustafa Raja: I would also want to see how it… how it’s performing on live questions.
34 00:04:51.640 ⇒ 00:04:55.829 Casie Aviles: Yeah, and also the latency, if we have to check the latency, if it.
35 00:04:56.300 ⇒ 00:04:58.599 Samuel Roberts: Yes. Good point, good point.
36 00:04:58.820 ⇒ 00:05:08.120 Samuel Roberts: Yeah, so if we can kind of confirm those things, that it’s, you know, on par with real-world questions, and then,
37 00:05:09.520 ⇒ 00:05:13.989 Samuel Roberts: Yeah, the latency, that we’d probably have to test a little bit more on our own, maybe.
38 00:05:14.250 ⇒ 00:05:15.180 Samuel Roberts: Because we want.
39 00:05:15.180 ⇒ 00:05:15.670 Mustafa Raja: I’m just…
40 00:05:15.670 ⇒ 00:05:19.230 Samuel Roberts: full, Full circle latency, the full loop.
41 00:05:19.700 ⇒ 00:05:20.360 Mustafa Raja: Yeah.
42 00:05:20.360 ⇒ 00:05:21.189 Amber Lin: I see.
43 00:05:21.190 ⇒ 00:05:21.650 Samuel Roberts: Okay.
44 00:05:21.650 ⇒ 00:05:26.439 Amber Lin: My also… another question is, I’m looking at the Gantt right now.
45 00:05:26.550 ⇒ 00:05:32.929 Amber Lin: it seems like we haven’t migrated everything, or we just haven’t updated the tickets yet. For example.
46 00:05:33.850 ⇒ 00:05:36.250 Amber Lin: It’s, the feedback.
47 00:05:36.930 ⇒ 00:05:43.769 Samuel Roberts: No, that’s true, that’s not there yet either. That actually does have to happen before, so that’s why it’s still in dev and not in, like, a staging and…
48 00:05:44.580 ⇒ 00:05:47.060 Samuel Roberts: Ready to go. So, that kind of got…
49 00:05:47.450 ⇒ 00:05:51.130 Samuel Roberts: back-seated for a little bit. Yeah.
50 00:05:51.130 ⇒ 00:05:52.110 Amber Lin: Let’s see…
51 00:05:52.800 ⇒ 00:05:54.399 Samuel Roberts: Thank you for that, yeah, I…
52 00:05:54.400 ⇒ 00:06:03.140 Amber Lin: Let’s… I’ll try… let me know if my audio starts breaking up. Okay. So let’s look at this right now.
53 00:06:03.360 ⇒ 00:06:06.010 Amber Lin: And these two are done, right?
54 00:06:06.260 ⇒ 00:06:06.930 Mustafa Raja: Yeah.
55 00:06:06.930 ⇒ 00:06:09.140 Samuel Roberts: Yeah, Cloud SQL is done, yep, both of those are done, yeah.
56 00:06:09.940 ⇒ 00:06:15.380 Amber Lin: Great. Start from the top.
57 00:06:16.140 ⇒ 00:06:20.580 Amber Lin: This probably is next week, I assume?
58 00:06:21.680 ⇒ 00:06:27.589 Mustafa Raja: I think… I think once we set up the live questions, that should be able to do that, no? What do you think, Sam?
59 00:06:29.730 ⇒ 00:06:31.570 Samuel Roberts: Sorry, which ones? I can’t…
60 00:06:31.570 ⇒ 00:06:32.870 Mustafa Raja: 20th.
61 00:06:33.040 ⇒ 00:06:34.409 Mustafa Raja: The 20th one.
62 00:06:34.980 ⇒ 00:06:41.580 Samuel Roberts: Yeah, once we… Right… oh, yeah, yeah, okay, yeah, I think…
63 00:06:41.860 ⇒ 00:06:44.820 Samuel Roberts: Once we flip that over, that should be a good way to test it, yeah.
64 00:06:45.500 ⇒ 00:06:50.200 Mustafa Raja: It might just not be concurrent, but it’s going to be something like that
65 00:06:50.780 ⇒ 00:06:54.620 Mustafa Raja: Yeah, exactly. Yeah, it’s true, that’s true. We could probably do a little bit more for concurrent, but that’s a good point.
66 00:06:55.600 ⇒ 00:06:56.650 Mustafa Raja: Yeah.
67 00:06:58.640 ⇒ 00:07:02.260 Amber Lin: Going down here, so we have…
68 00:07:02.430 ⇒ 00:07:08.949 Amber Lin: Oh, by the way, and then we have feedback, capture, and logging. So this is for the triage…
69 00:07:09.160 ⇒ 00:07:10.880 Amber Lin: triage tickets.
70 00:07:11.160 ⇒ 00:07:14.679 Amber Lin: So, how does… how are these looking?
71 00:07:15.370 ⇒ 00:07:18.740 Casie Aviles: These haven’t been started at all.
72 00:07:25.400 ⇒ 00:07:26.070 Casie Aviles: I don’t think…
73 00:07:26.070 ⇒ 00:07:28.380 Samuel Roberts: How… Yeah, go ahead.
74 00:07:28.910 ⇒ 00:07:31.500 Casie Aviles: Yeah, I was just thinking if,
75 00:07:32.480 ⇒ 00:07:35.459 Casie Aviles: If this needs to happen first, or…
76 00:07:37.410 ⇒ 00:07:41.220 Casie Aviles: If there’s, something else that needs to happen before this…
77 00:07:41.890 ⇒ 00:07:56.080 Amber Lin: I was thinking that if we just want to test if the rag and the prompt works the same, we don’t necessarily need the feedback, but that’s the only way if we’re testing with CSRs and not just with ROs.
78 00:07:56.080 ⇒ 00:07:57.230 Samuel Roberts: Yeah.
79 00:07:57.390 ⇒ 00:07:57.860 Amber Lin: they need.
80 00:07:57.860 ⇒ 00:07:59.040 Samuel Roberts: Yeah, no, I think…
81 00:07:59.610 ⇒ 00:08:09.999 Samuel Roberts: Totally. So before we roll it out to them, that has… the feedback definitely needs to get done, but for just testing what they’re asking into Master to make sure it’s ready to roll out that way, I think is…
82 00:08:10.760 ⇒ 00:08:11.730 Samuel Roberts: It’s fine.
83 00:08:14.020 ⇒ 00:08:14.670 Amber Lin: Yeah.
84 00:08:14.840 ⇒ 00:08:30.920 Amber Lin: Okay, so… I think we’re mainly looking at this section right here. We have these tickets, we have the migrate, we have the real, the zip code, and then the staging-related stuff. How should we adjust the timeline?
85 00:08:37.220 ⇒ 00:08:40.900 Samuel Roberts: I would say… Bruce.
86 00:08:40.909 ⇒ 00:08:43.890 Casie Aviles: Probably less of a priority.
87 00:08:44.720 ⇒ 00:08:45.640 Amber Lin: Which one?
88 00:08:45.760 ⇒ 00:08:48.970 Casie Aviles: For real, I don’t think that’s a high priority.
89 00:08:49.210 ⇒ 00:08:59.580 Amber Lin: Okay, so that’s for… When we start using the new, like, the staging, does the…
90 00:08:59.900 ⇒ 00:09:03.029 Amber Lin: all the logs get logged in BigQuery.
91 00:09:03.710 ⇒ 00:09:07.270 Mustafa Raja: We have one single project for BigQuery, so it’s going to go in there.
92 00:09:07.990 ⇒ 00:09:09.599 Amber Lin: Oh, I see, so in order.
93 00:09:09.600 ⇒ 00:09:10.770 Mustafa Raja: We could have different…
94 00:09:10.770 ⇒ 00:09:12.599 Amber Lin: new usage, we will.
95 00:09:12.600 ⇒ 00:09:15.989 Mustafa Raja: Yeah, we could have, yeah, yeah. What we could do is…
96 00:09:15.990 ⇒ 00:09:19.700 Samuel Roberts: It’s being logged, though, to the postcard… or the Cloud SQL, right?
97 00:09:19.700 ⇒ 00:09:29.259 Mustafa Raja: Yeah, yeah, what we could do is, we could, create multiple datasets for production, staging, and dev. That way, we could segregate the data.
98 00:09:30.570 ⇒ 00:09:31.720 Samuel Roberts: That’s probably smart.
99 00:09:31.830 ⇒ 00:09:36.580 Samuel Roberts: Right now, we have just the one…
100 00:09:36.580 ⇒ 00:09:49.630 Mustafa Raja: Yeah, yeah. Cloud SQL with the one master, right? Yeah, yeah, one master, and then one BigQuery, right? And I think, they wouldn’t want to add more BigQuery projects, since we can’t just create datasets, right? So…
101 00:09:52.390 ⇒ 00:10:06.080 Amber Lin: I see. So, to summarize, right now, even when we do start the staging and start the new environment, logs get pointed to BigQuery, but our real is based on Snowflake, so we will need to…
102 00:10:06.080 ⇒ 00:10:11.729 Mustafa Raja: logs are right now in Cloud SQL. We just need to write a small function to move them to…
103 00:10:11.840 ⇒ 00:10:13.440 Mustafa Raja: Gotcha. And I’ll be creating…
104 00:10:14.140 ⇒ 00:10:14.830 Amber Lin: Okay.
105 00:10:14.830 ⇒ 00:10:18.660 Mustafa Raja: So… We just need to format it better, and then push it. That’s all.
106 00:10:18.970 ⇒ 00:10:22.489 Amber Lin: Okay, sounds good. So I’m gonna move this to…
107 00:10:22.650 ⇒ 00:10:27.020 Amber Lin: before this, I think I’ll push this back a week.
108 00:10:27.020 ⇒ 00:10:27.710 Samuel Roberts: Okay.
109 00:10:28.250 ⇒ 00:10:34.300 Amber Lin: And then push this to next week, because I don’t think we’ll be ready, and we don’t…
110 00:10:34.300 ⇒ 00:10:34.810 Samuel Roberts: It’s like…
111 00:10:34.810 ⇒ 00:10:36.389 Amber Lin: really lower priority.
112 00:10:36.850 ⇒ 00:10:37.910 Samuel Roberts: Okay. Okay.
113 00:10:38.060 ⇒ 00:10:41.970 Amber Lin: Okay. Then looking at…
114 00:10:47.050 ⇒ 00:10:51.230 Amber Lin: these… How are these looking?
115 00:10:51.910 ⇒ 00:10:56.800 Amber Lin: When will we… are we doing this first, and then doing the staging next week?
116 00:10:56.800 ⇒ 00:10:58.609 Samuel Roberts: Yes, staging should move.
117 00:10:58.800 ⇒ 00:10:59.840 Samuel Roberts: Yeah.
118 00:11:00.270 ⇒ 00:11:01.100 Amber Lin: Okay.
119 00:11:01.760 ⇒ 00:11:03.039 Amber Lin: Okay, I probably…
120 00:11:03.040 ⇒ 00:11:03.880 Samuel Roberts: Yep, yep.
121 00:11:03.880 ⇒ 00:11:10.540 Amber Lin: blog, we should… And then this one… What about the migrate app?
122 00:11:10.920 ⇒ 00:11:11.689 Amber Lin: What about the…
123 00:11:11.690 ⇒ 00:11:19.950 Mustafa Raja: I think what we discussed is we need, I need to write up a small document that we can present to them.
124 00:11:20.350 ⇒ 00:11:22.280 Mustafa Raja: I need to redo that.
125 00:11:22.430 ⇒ 00:11:23.290 Mustafa Raja: Yeah.
126 00:11:23.290 ⇒ 00:11:24.030 Amber Lin: Gotcha, okay.
127 00:11:24.030 ⇒ 00:11:28.599 Mustafa Raja: Let’s ticket this out, so I know that it’s there, I just didn’t know that it…
128 00:11:28.600 ⇒ 00:11:34.239 Amber Lin: Yeah, that’s a good point. I think… I just… I have the ticket, I don’t know if I put it in cycle.
129 00:11:34.240 ⇒ 00:11:35.559 Mustafa Raja: Oh, okay, okay.
130 00:11:36.290 ⇒ 00:11:40.640 Amber Lin: Yeah, this one.
131 00:11:42.640 ⇒ 00:11:46.020 Amber Lin: So I’m gonna move the staging to next week.
132 00:11:46.540 ⇒ 00:11:50.880 Amber Lin: Cycle…
133 00:11:54.490 ⇒ 00:11:57.100 Amber Lin: Cycle to next week.
134 00:11:58.000 ⇒ 00:12:02.350 Amber Lin: Eval… is this done? This is done, right?
135 00:12:02.820 ⇒ 00:12:07.699 Mustafa Raja: Yes, but I need to groom this ticket. I’ll mark it as done once I do that.
136 00:12:07.700 ⇒ 00:12:09.339 Amber Lin: Okay, sounds good.
137 00:12:09.490 ⇒ 00:12:11.629 Amber Lin: Alright.
138 00:12:13.670 ⇒ 00:12:15.790 Amber Lin: What about…
139 00:12:16.030 ⇒ 00:12:23.659 Amber Lin: So we’ll do the… do the plan, and hopefully do it next week, or maybe later. What about these?
140 00:12:24.570 ⇒ 00:12:29.140 Amber Lin: a table for zip database migration. Do we need that?
141 00:12:31.100 ⇒ 00:12:42.719 Casie Aviles: Right now they’re in Superbase, so if we have Cloud SQL, then we’ll need to move everything we have from Superbase to.
142 00:12:43.160 ⇒ 00:12:47.489 Samuel Roberts: I think that’s probably fine to delay longer, though, because we can still use Supabase even if we move over.
143 00:12:47.490 ⇒ 00:12:48.010 Mustafa Raja: Hmm.
144 00:12:48.010 ⇒ 00:12:48.780 Samuel Roberts: Indy.
145 00:12:48.970 ⇒ 00:12:52.109 Amber Lin: I see, so this is the same thing as the rail stuff.
146 00:12:52.620 ⇒ 00:12:53.500 Amber Lin: Okay.
147 00:12:54.660 ⇒ 00:13:00.689 Casie Aviles: Yay… I’m not sure if there are any immediate benefits to moving it right now.
148 00:13:00.750 ⇒ 00:13:01.260 Mustafa Raja: Yeah.
149 00:13:01.260 ⇒ 00:13:03.780 Samuel Roberts: No, I think it’s just getting it onto their stuff eventually.
150 00:13:04.000 ⇒ 00:13:04.380 Casie Aviles: Okay.
151 00:13:04.730 ⇒ 00:13:11.840 Amber Lin: I’ll say this is to Cloud SQL, and then same for… same for this one.
152 00:13:17.080 ⇒ 00:13:18.710 Amber Lin: Cool. Alright.
153 00:13:19.010 ⇒ 00:13:22.619 Amber Lin: So, that will be next week. What about…
154 00:13:23.070 ⇒ 00:13:26.050 Mustafa Raja: 46 is a… is a duplicate.
155 00:13:26.160 ⇒ 00:13:29.620 Amber Lin: I am gonna delete this one.
156 00:13:32.400 ⇒ 00:13:35.950 Amber Lin: What about the Google Chat web book?
157 00:13:36.650 ⇒ 00:13:40.500 Mustafa Raja: Yeah, I think for DevOne, it’s in place, right? Right, Casey?
158 00:13:41.490 ⇒ 00:13:50.360 Casie Aviles: Yeah, actually, the only ones that it won’t be pointing to are, like, the… Wait, I have to…
159 00:13:50.460 ⇒ 00:13:51.600 Casie Aviles: Let me draw.
160 00:13:51.970 ⇒ 00:13:55.360 Casie Aviles: These ones are separate endpoints.
161 00:13:55.660 ⇒ 00:14:00.259 Casie Aviles: I see. But I guess that’s separate, since it’s only Mastra, so I think…
162 00:14:00.260 ⇒ 00:14:00.710 Amber Lin: That’s…
163 00:14:00.710 ⇒ 00:14:01.800 Casie Aviles: This is fine. Okay.
164 00:14:02.370 ⇒ 00:14:06.210 Amber Lin: So I’ll just say, I’ll just say this is done, yeah.
165 00:14:07.390 ⇒ 00:14:08.640 Amber Lin: Okay, great.
166 00:14:12.320 ⇒ 00:14:14.090 Amber Lin: Alright, so…
167 00:14:14.410 ⇒ 00:14:20.660 Amber Lin: Sounds like this week, our may… I mean, this one, we’re also not… are we doing this one? This week?
168 00:14:21.600 ⇒ 00:14:22.650 Amber Lin: embedding pipeline?
169 00:14:22.650 ⇒ 00:14:23.670 Samuel Roberts: replicate.
170 00:14:25.990 ⇒ 00:14:26.660 Mustafa Raja: I couldn’t…
171 00:14:27.910 ⇒ 00:14:29.780 Samuel Roberts: No, I’d say… I’d say bump that.
172 00:14:30.040 ⇒ 00:14:36.530 Samuel Roberts: Personally. I think, I think I’d rather worry about getting Andy actually moved over, and then we can move that over.
173 00:14:36.710 ⇒ 00:14:44.070 Amber Lin: I see. So this… we would… I think we would do after we give them access, or is this blocking anything?
174 00:14:44.380 ⇒ 00:14:48.189 Samuel Roberts: I don’t think it’s blocking anything except for just fully depreciating. Okay.
175 00:14:48.190 ⇒ 00:14:49.710 Mustafa Raja: Yeah, yeah, yeah.
176 00:14:49.710 ⇒ 00:14:51.850 Amber Lin: Just gonna put it over…
177 00:14:51.850 ⇒ 00:14:52.890 Samuel Roberts: That’s fine, yeah.
178 00:14:52.890 ⇒ 00:14:53.860 Amber Lin: Oh, we’re here.
179 00:14:54.860 ⇒ 00:14:55.590 Amber Lin: Okay.
180 00:15:06.470 ⇒ 00:15:15.730 Amber Lin: Whatever. Okay, so this week, I think our main… Our main task is…
181 00:15:15.970 ⇒ 00:15:20.669 Amber Lin: Migrate… oh, by the way, and make sure the triage is done.
182 00:15:24.670 ⇒ 00:15:27.500 Casie Aviles: For this V. This one… these ones, then.
183 00:15:28.030 ⇒ 00:15:35.580 Amber Lin: Right, it… I don’t… I don’t know if we’ll be on you. I guess after we talk about the central dog stuff, we can see…
184 00:15:35.780 ⇒ 00:15:37.260 Amber Lin: Who has time?
185 00:15:37.850 ⇒ 00:15:42.189 Amber Lin: Can I merge this ticket into the triage? Because it’s… it’s the…
186 00:15:42.610 ⇒ 00:15:47.300 Amber Lin: Like, it’s thumbs up, then down, and then the triage, they’re… they’re kind of…
187 00:15:47.370 ⇒ 00:15:51.510 Samuel Roberts: Wait, what is it? Testing of enriched responses.
188 00:15:51.920 ⇒ 00:15:53.350 Samuel Roberts: This is just… Yeah, I think that’s just…
189 00:15:53.350 ⇒ 00:15:53.920 Amber Lin: Or they…
190 00:15:53.920 ⇒ 00:15:56.060 Samuel Roberts: Okay, I think it’s just validating, yeah.
191 00:15:56.260 ⇒ 00:15:58.489 Amber Lin: I think it’s fine. Yeah, I’m gonna delete this.
192 00:15:59.990 ⇒ 00:16:00.450 Samuel Roberts: That’s fine.
193 00:16:00.450 ⇒ 00:16:04.470 Amber Lin: and say… I’ll say thumbs up and down.
194 00:16:16.460 ⇒ 00:16:23.249 Amber Lin: Okay, so I think this will include thumbs up, thumbs down, the text block they send.
195 00:16:24.820 ⇒ 00:16:28.559 Amber Lin: Thumbs up and down button.
196 00:16:28.680 ⇒ 00:16:33.129 Amber Lin: To feedback text box.
197 00:16:33.550 ⇒ 00:16:36.690 Amber Lin: All three logging in here.
198 00:16:37.710 ⇒ 00:16:38.920 Amber Lin: issue.
199 00:16:39.610 ⇒ 00:16:40.420 Amber Lin: Okay.
200 00:16:41.470 ⇒ 00:16:43.480 Amber Lin: Let’s see… Alright.
201 00:16:44.070 ⇒ 00:16:49.110 Amber Lin: Okay, so I guess this week, our main goal is… These two…
202 00:16:49.730 ⇒ 00:16:55.070 Amber Lin: And then the text to SQL… Stop.
203 00:16:55.070 ⇒ 00:16:59.000 Casie Aviles: Yeah, I’m… I’m pushing to finish the zip code stuff.
204 00:16:59.430 ⇒ 00:17:00.779 Amber Lin: Okay, I’m… This week.
205 00:17:01.030 ⇒ 00:17:06.710 Amber Lin: Put this down to the zip code, so we know… that.
206 00:17:07.510 ⇒ 00:17:09.539 Amber Lin: Oh god, it’s so slow.
207 00:17:10.040 ⇒ 00:17:12.540 Amber Lin: So that we know they’re together.
208 00:17:13.750 ⇒ 00:17:14.750 Amber Lin: Okay.
209 00:17:18.650 ⇒ 00:17:27.529 Amber Lin: Okay, sounds good. We’ll make sure I can go look at the tickets in a bit. Let’s talk about the Central Doc stuff.
210 00:17:31.540 ⇒ 00:17:38.569 Amber Lin: Is this still the plan we’re going with? And I don’t think we’re… we’re on track for the timeline, so…
211 00:17:38.570 ⇒ 00:17:40.110 Samuel Roberts: Definitely not on track, yeah, I don’t…
212 00:17:40.110 ⇒ 00:17:40.730 Amber Lin: Yeah.
213 00:17:40.730 ⇒ 00:17:45.110 Samuel Roberts: I mean, honestly, when we were digging through some of the contradictory and duplicate stuff, I don’t know.
214 00:17:46.570 ⇒ 00:17:53.709 Samuel Roberts: I don’t know if all of this makes… makes a ton of… like, certain things make sense, but, like, other things might not anymore, so…
215 00:17:54.200 ⇒ 00:17:54.860 Amber Lin: Yeah, yeah.
216 00:17:54.990 ⇒ 00:18:12.150 Amber Lin: I mean, we can start off with this. I still think there’s more in the questions categorization to explore, but luckily, it did say that the zip code stuff was the main, like, the biggest category asked, so I’m glad that we spent a lot of time working on.
217 00:18:12.150 ⇒ 00:18:13.900 Samuel Roberts: Yeah, definitely.
218 00:18:14.000 ⇒ 00:18:23.420 Amber Lin: So, I still think this one is a priority, but after that, you’re right, like, I just don’t know if our plan is still valid.
219 00:18:23.840 ⇒ 00:18:29.260 Samuel Roberts: Yeah, I mean, it’s just, it’s such a… meaty problem, you know?
220 00:18:29.260 ⇒ 00:18:29.860 Amber Lin: Hmm.
221 00:18:30.400 ⇒ 00:18:32.400 Samuel Roberts: And it’s not just a…
222 00:18:32.400 ⇒ 00:18:33.270 Amber Lin: Down.
223 00:18:33.270 ⇒ 00:18:37.130 Samuel Roberts: Yeah, that’s what I’m trying to… I mean, we tried to think that through a little bit, but I think it didn’t…
224 00:18:37.530 ⇒ 00:18:39.349 Samuel Roberts: quite match.
225 00:18:39.350 ⇒ 00:18:43.409 Amber Lin: Let’s pull up, let’s pull up this…
226 00:18:43.940 ⇒ 00:18:55.099 Amber Lin: initial report, like, we’re working on this issue. Yeah, sure. And so we’re able to figure that one out, and that covers, I would say, half or 40% of the questions.
227 00:18:55.200 ⇒ 00:18:58.579 Amber Lin: Let’s… we can just go by the…
228 00:18:58.990 ⇒ 00:19:03.240 Amber Lin: Order of how many questions are asked, and tackle it that way.
229 00:19:06.170 ⇒ 00:19:07.319 Amber Lin: Like this one.
230 00:19:07.320 ⇒ 00:19:07.900 Samuel Roberts: because…
231 00:19:07.900 ⇒ 00:19:12.070 Amber Lin: This should be an easier issue to tackle anyways.
232 00:19:15.410 ⇒ 00:19:27.109 Samuel Roberts: Right, right. Yeah, because, like, that was one thing that we saw when, Mustafa and I went through the duplicate stuff, was all the different times, and all the different prices.
233 00:19:27.230 ⇒ 00:19:36.280 Samuel Roberts: And I think what we probably realistically need is not just, like, deduplication, but a better structured format for them to…
234 00:19:36.280 ⇒ 00:19:36.840 Amber Lin: I agree.
235 00:19:36.840 ⇒ 00:19:38.150 Samuel Roberts: Be in the document.
236 00:19:38.570 ⇒ 00:19:39.280 Amber Lin: Yes.
237 00:19:39.920 ⇒ 00:19:59.549 Amber Lin: That’s what I wanted, I think, from the start, is that I think we’ll need to give them a structure. We’ll fill in what we have, and then they need to go in and validate each one. But I’m glad we also did the duplication, because then we know, kind of, what type of information is there.
238 00:19:59.560 ⇒ 00:20:08.810 Amber Lin: I wanted to ask Pranav to get sample questions, but I don’t think the new report he made
239 00:20:08.920 ⇒ 00:20:12.010 Amber Lin: Had those, so that was my ask.
240 00:20:12.120 ⇒ 00:20:19.580 Amber Lin: For him, so that we know what we’re dealing with, because I think if we were to pass those…
241 00:20:19.940 ⇒ 00:20:25.409 Amber Lin: like, some sample of these questions in… into the AI, then it can help us
242 00:20:25.650 ⇒ 00:20:29.000 Amber Lin: Figure out what information we’re looking for.
243 00:20:36.240 ⇒ 00:20:37.050 Samuel Roberts: Okay.
244 00:20:37.960 ⇒ 00:20:39.650 Amber Lin: Yeah.
245 00:20:39.650 ⇒ 00:20:44.010 Samuel Roberts: Yeah, I… yeah, I don’t know, I’m sorry, this is just, like, I’m trying to think through what we’re talking about.
246 00:20:44.010 ⇒ 00:20:46.840 Amber Lin: Yeah, makes sense. This is a big problem, like, there’s a very.
247 00:20:46.840 ⇒ 00:20:47.430 Samuel Roberts: a big issue.
248 00:20:47.430 ⇒ 00:20:51.319 Amber Lin: That we’re trying to break down, but this is just a suggestion that we can go.
249 00:20:51.320 ⇒ 00:20:53.420 Samuel Roberts: Yeah, I think that’s probably a good way to do it.
250 00:20:53.550 ⇒ 00:20:58.650 Samuel Roberts: Because, yeah, I mean, it’s the same thing with pricing costs, it’s the same thing with scheduling, it’s the same thing with…
251 00:20:58.650 ⇒ 00:20:59.270 Amber Lin: Yeah.
252 00:20:59.390 ⇒ 00:21:01.730 Samuel Roberts: Some of the procedures, yeah.
253 00:21:02.880 ⇒ 00:21:04.770 Samuel Roberts: Definition of coverage, yeah, I think…
254 00:21:05.230 ⇒ 00:21:13.359 Amber Lin: Would you say it’s… it’s better if we… I would say, like, we can probably get… do these two later, that can.
255 00:21:13.360 ⇒ 00:21:14.000 Samuel Roberts: Yeah, I would kick.
256 00:21:14.000 ⇒ 00:21:15.859 Amber Lin: That’s for runtime.
257 00:21:16.390 ⇒ 00:21:21.740 Samuel Roberts: Definitely. I think that’s fine to pump those down. I think something like… the…
258 00:21:26.310 ⇒ 00:21:31.329 Samuel Roberts: the, like, SOP auto formatter, or at least the basics of that.
259 00:21:31.330 ⇒ 00:21:35.670 Amber Lin: Do you think we should give them, like, the structure first?
260 00:21:35.670 ⇒ 00:21:40.820 Samuel Roberts: Well, that’s what I was gonna say, I feel like the Central Doc Policy Formatting and Structure Guidelines is really, like, this is probably…
261 00:21:41.980 ⇒ 00:21:46.650 Amber Lin: Okay, so let’s… let’s move… Let’s move that.
262 00:21:46.760 ⇒ 00:21:54.660 Amber Lin: Here, so let’s… I’m just gonna move this to this section.
263 00:21:54.900 ⇒ 00:22:02.020 Amber Lin: We do the weekly report, we do this… initial audit, I’ll call it.
264 00:22:02.310 ⇒ 00:22:03.670 Amber Lin: More content.
265 00:22:03.930 ⇒ 00:22:05.130 Amber Lin: Audit.
266 00:22:05.860 ⇒ 00:22:15.759 Amber Lin: And then… Let’s say we… We do this…
267 00:22:19.050 ⇒ 00:22:24.110 Amber Lin: Recommended structure… .
268 00:22:27.710 ⇒ 00:22:31.539 Mustafa Raja: If it’s dependent, we’re gonna move it before that, right? Yeah.
269 00:22:32.010 ⇒ 00:22:32.750 Amber Lin: Oops.
270 00:22:33.060 ⇒ 00:22:38.030 Amber Lin: Dependencies… There we go.
271 00:22:42.450 ⇒ 00:22:43.240 Amber Lin: Okay.
272 00:22:43.790 ⇒ 00:22:57.609 Amber Lin: So… Oh, God, this is so messy. So we do that one, I’m gonna copy… this table…
273 00:22:57.970 ⇒ 00:22:59.560 Amber Lin: Okay, I can’t do that.
274 00:23:00.910 ⇒ 00:23:13.390 Amber Lin: So what’s next? So, if we… if we give them a policy, give them structure, what are we going to do next?
275 00:23:14.170 ⇒ 00:23:15.460 Amber Lin: after that.
276 00:23:17.590 ⇒ 00:23:25.489 Samuel Roberts: Well, I think we have to kind of figure that structure out first, more than anything. I don’t think it’s obvious what that needs to be.
277 00:23:26.260 ⇒ 00:23:36.299 Amber Lin: Okay, do we have a rough idea of what it needs to be? Because we know, like, do we want it in table format? Tabular format?
278 00:23:36.310 ⇒ 00:23:51.000 Amber Lin: How many of their data is actually, like, procedural and need to be lines or paragraphs of text, versus some things are so concrete, it just this equals this, that equals that, like, would that make it easier?
279 00:23:53.380 ⇒ 00:23:58.579 Samuel Roberts: it might make it easier in terms of consolidating it, but I don’t know how that would affect the…
280 00:23:58.830 ⇒ 00:24:01.260 Samuel Roberts: Embedding, necessarily, either, though.
281 00:24:04.970 ⇒ 00:24:11.960 Amber Lin: Yeah, why don’t we… Okay, these are evals…
282 00:24:17.650 ⇒ 00:24:18.600 Amber Lin: Okay.
283 00:24:19.230 ⇒ 00:24:23.260 Amber Lin: Could we start a Notion doc and talk about that?
284 00:24:25.000 ⇒ 00:24:27.810 Amber Lin: And just start off with a preliminary.
285 00:24:27.810 ⇒ 00:24:28.180 Samuel Roberts: Yeah.
286 00:24:28.440 ⇒ 00:24:30.339 Amber Lin: Structure of how it should look?
287 00:24:30.760 ⇒ 00:24:34.670 Samuel Roberts: Probably not a bad idea. Like, even if we just think of it as, like, an idealized, what is it.
288 00:24:34.670 ⇒ 00:24:35.300 Amber Lin: Yeah, yeah.
289 00:24:36.180 ⇒ 00:24:40.340 Samuel Roberts: And then start mapping things to it from the document, that might not be a bad way.
290 00:24:40.590 ⇒ 00:24:47.610 Amber Lin: Yeah, we can generate some, and we can also ask AI to give some recommended, like, document structure.
291 00:24:47.710 ⇒ 00:24:54.310 Amber Lin: Where should we… okay, let me… Find Notion.
292 00:25:01.940 ⇒ 00:25:08.299 Casie Aviles: Yeah, maybe an A-B test of, like, the old structure and the new structure could also help, like…
293 00:25:10.110 ⇒ 00:25:13.690 Casie Aviles: think about… Whether this structure is effective or not.
294 00:25:14.490 ⇒ 00:25:15.180 Samuel Roberts: Yeah.
295 00:25:20.790 ⇒ 00:25:24.936 Amber Lin: Cool. I think…
296 00:25:39.450 ⇒ 00:25:44.439 Amber Lin: Okay, my computer’s really slow. I’m gonna say we look at the…
297 00:25:44.560 ⇒ 00:25:49.060 Amber Lin: Just put it in the bottom of the central dock improvements.
298 00:25:50.400 ⇒ 00:25:51.360 Amber Lin: Doc.
299 00:25:55.830 ⇒ 00:25:58.700 Amber Lin: The one we were working on before with the tongue?
300 00:26:15.940 ⇒ 00:26:19.670 Mustafa Raja: But if you’re having difficulty sharing the screen, I could share the screen.
301 00:26:19.670 ⇒ 00:26:25.140 Amber Lin: Yeah, that would be great. It’s, it’s called Central Dock Improvement.
302 00:26:25.140 ⇒ 00:26:26.709 Samuel Roberts: Okay, yeah, I was pulling it up, too.
303 00:26:28.000 ⇒ 00:26:30.089 Mustafa Raja: Where would be that?
304 00:26:30.320 ⇒ 00:26:31.120 Mustafa Raja: Sorry.
305 00:26:31.680 ⇒ 00:26:32.200 Samuel Roberts: In no space.
306 00:26:32.200 ⇒ 00:26:36.389 Amber Lin: If you search ABC Central Dock Improvement.
307 00:26:37.240 ⇒ 00:26:38.460 Mustafa Raja: Let me see…
308 00:26:39.860 ⇒ 00:26:42.330 Samuel Roberts: It’s in Notion, you’re talking about Notion.
309 00:26:42.330 ⇒ 00:26:43.080 Mustafa Raja: Hi.
310 00:26:46.530 ⇒ 00:26:47.190 Samuel Roberts: Yeah.
311 00:26:53.830 ⇒ 00:26:57.300 Mustafa Raja: Abc… sorry.
312 00:26:57.470 ⇒ 00:27:02.730 Mustafa Raja: In the season… Central… okay, central lock improvement, right?
313 00:27:04.450 ⇒ 00:27:05.040 Amber Lin: Yes.
314 00:27:05.730 ⇒ 00:27:09.049 Mustafa Raja: And then… what do you want to add at the end?
315 00:27:09.850 ⇒ 00:27:11.359 Amber Lin: Yeah, at the end, please.
316 00:27:11.840 ⇒ 00:27:14.120 Mustafa Raja: Yeah, yeah, what did you want to add, sorry?
317 00:27:15.050 ⇒ 00:27:18.940 Amber Lin: proposed structure…
318 00:27:22.030 ⇒ 00:27:24.890 Amber Lin: Alright, let’s… let’s write together.
319 00:27:26.360 ⇒ 00:27:30.870 Amber Lin: Oh… What would be… what would we…
320 00:27:31.510 ⇒ 00:27:39.260 Amber Lin: start with, like, I have an idea, but my idea is based on, in the past, how I work with pests.
321 00:27:39.400 ⇒ 00:27:46.589 Amber Lin: Essentially, it just… they have SOPs, they have service informations,
322 00:27:46.720 ⇒ 00:27:53.170 Amber Lin: And then, like, service information would include, like, what the service is, and…
323 00:27:53.420 ⇒ 00:28:02.800 Amber Lin: price and timing, so I guess there’s SOP… Service… specific information.
324 00:28:02.970 ⇒ 00:28:22.910 Amber Lin: What the service is… A verge… Pricing, timing… Coverage… personnel, even?
325 00:28:25.000 ⇒ 00:28:27.230 Samuel Roberts: Personnel worries me, because I don’t know how that…
326 00:28:27.230 ⇒ 00:28:28.170 Amber Lin: Yeah.
327 00:28:28.170 ⇒ 00:28:30.040 Samuel Roberts: The zip code, and that…
328 00:28:30.040 ⇒ 00:28:30.640 Amber Lin: Yeah.
329 00:28:30.640 ⇒ 00:28:31.650 Samuel Roberts: stale.
330 00:28:32.750 ⇒ 00:28:35.979 Amber Lin: I guess they have it in the dock right now, we can find a way…
331 00:28:36.720 ⇒ 00:28:38.939 Samuel Roberts: Yeah, no, it’s definitely in there right now, you’re right.
332 00:28:43.910 ⇒ 00:28:46.739 Amber Lin: Okay, and then they have shared…
333 00:28:46.920 ⇒ 00:29:05.469 Amber Lin: SOP info, which is, like, billing, cancellations… And I guess that… But… basic SOPs, no.
334 00:29:06.510 ⇒ 00:29:16.300 Amber Lin: Scheduling an estimate… Scheduling… Yeah.
335 00:29:28.650 ⇒ 00:29:31.510 Amber Lin: Okay, I mean, this is what I can think of.
336 00:29:31.770 ⇒ 00:29:39.609 Amber Lin: And I know when we were looking at it, there was a lot of, like, coverage issues, pricing overlap, timing overlap.
337 00:29:39.760 ⇒ 00:29:41.590 Amber Lin: We also didn’t…
338 00:29:41.590 ⇒ 00:29:42.190 Samuel Roberts: But I mean…
339 00:29:43.060 ⇒ 00:29:45.150 Samuel Roberts: right track here, though. Yeah, I…
340 00:29:45.350 ⇒ 00:29:54.979 Samuel Roberts: I think you have the better… you have the best sense of, like, what the overall structure kind of wants to be, but isn’t quite, and so I think this is probably a good starting point.
341 00:29:55.680 ⇒ 00:29:56.510 Amber Lin: Yeah.
342 00:29:57.210 ⇒ 00:30:02.069 Amber Lin: This is kind of how I try to organize the past documents.
343 00:30:02.070 ⇒ 00:30:02.650 Samuel Roberts: Okay.
344 00:30:02.870 ⇒ 00:30:08.689 Amber Lin: But then there’s still a lot of… like, the dog is still pretty long. I guess.
345 00:30:08.690 ⇒ 00:30:09.600 Samuel Roberts: Is my caution?
346 00:30:09.600 ⇒ 00:30:22.260 Amber Lin: I can give my general sense of what type of information is in there, but I also don’t know how best to format it for RAG, for chunking, and how.
347 00:30:22.260 ⇒ 00:30:22.780 Mustafa Raja: Hmm.
348 00:30:23.390 ⇒ 00:30:32.910 Amber Lin: section things so that it best comes up in the answers, because we can either, say, have a unique SOP,
349 00:30:32.910 ⇒ 00:30:44.929 Amber Lin: with specific service, specific info for each service, or we can have a general SOP, but then have it point out to different services, but when we chunk it, is it gonna work, or is it.
350 00:30:44.930 ⇒ 00:30:48.320 Samuel Roberts: Yeah, that’s… I’ll probably have to test that a little bit more, yeah.
351 00:30:48.320 ⇒ 00:30:50.120 Mustafa Raja: Yeah. Yeah, for that.
352 00:30:50.380 ⇒ 00:31:01.940 Mustafa Raja: I’m wondering how this shared SOP… so this is going to be a separate document that’s going to be shared among the different central docs, right? Am I correct?
353 00:31:02.830 ⇒ 00:31:04.079 Amber Lin: I thought there was a section…
354 00:31:04.080 ⇒ 00:31:04.770 Samuel Roberts: In the central…
355 00:31:04.770 ⇒ 00:31:18.460 Amber Lin: It’s currently a section, but if we’re combining all the rags, we might as well just have, like, a separate section. Because no other department other than pests reads the central dog, because the pest is the earliest one.
356 00:31:19.170 ⇒ 00:31:19.880 Samuel Roberts: Okay.
357 00:31:22.450 ⇒ 00:31:39.240 Mustafa Raja: Yeah, I’m asking because I’m wondering how the rag is going to recline, you know? How we, yeah, once we refer, I guess we could have… I guess we could add IDs, and then based on those IDs, we could, as we are…
358 00:31:39.530 ⇒ 00:31:44.639 Mustafa Raja: as we are embedding, we could replace the text with the IDs, you know, the shared text.
359 00:31:45.720 ⇒ 00:31:48.910 Mustafa Raja: But I guess we would want to… we would want to talk.
360 00:31:48.910 ⇒ 00:32:00.320 Amber Lin: I mean, we can always just start with copy and pasting the same thing in each central doc. Like, it will just be a duplicate, but I think it will be a shorter, short-term solution.
361 00:32:00.320 ⇒ 00:32:12.650 Mustafa Raja: Okay, okay. Okay, so the original sense of this shared, SOP would be, if, if this… if different sections within the same central dock have the same information, is that correct?
362 00:32:13.160 ⇒ 00:32:17.840 Amber Lin: Yeah, their billing info, cancellation info is essentially the same.
363 00:32:18.590 ⇒ 00:32:19.430 Mustafa Raja: Okay, okay.
364 00:32:22.710 ⇒ 00:32:25.489 Samuel Roberts: So it might be something like, you know, shared and, like, which…
365 00:32:26.900 ⇒ 00:32:29.420 Samuel Roberts: Departments it applies to, and then we can.
366 00:32:29.420 ⇒ 00:32:30.100 Mustafa Raja: use that.
367 00:32:30.910 ⇒ 00:32:31.690 Samuel Roberts: Ted.
368 00:32:31.690 ⇒ 00:32:33.370 Mustafa Raja: I guess? Like, rag it out.
369 00:32:33.370 ⇒ 00:32:34.690 Samuel Roberts: differently, embedded.
370 00:32:34.690 ⇒ 00:32:35.340 Mustafa Raja: Yeah.
371 00:32:35.440 ⇒ 00:32:41.010 Mustafa Raja: Yeah, I guess, I guess once we… once we, nail.
372 00:32:41.580 ⇒ 00:32:47.289 Samuel Roberts: this shared thing within a central dock, we could make it across multiple central docks, right?
373 00:32:50.850 ⇒ 00:32:51.690 Samuel Roberts: Yeah.
374 00:32:54.290 ⇒ 00:32:56.830 Samuel Roberts: Hmm, that’s interesting now that I’m thinking about it. I wonder…
375 00:32:58.240 ⇒ 00:32:59.060 Mustafa Raja: Yep.
376 00:33:05.430 ⇒ 00:33:12.010 Mustafa Raja: So the deduplication would most likely be for us, rather than for ag, right?
377 00:33:12.910 ⇒ 00:33:13.570 Mustafa Raja: Biggest down…
378 00:33:13.570 ⇒ 00:33:13.930 Samuel Roberts: Yeah.
379 00:33:13.930 ⇒ 00:33:14.249 Mustafa Raja: I think so.
380 00:33:14.250 ⇒ 00:33:18.700 Samuel Roberts: So, and I’m thinking maybe when we embed it, we just, yeah, we have it embed…
381 00:33:18.920 ⇒ 00:33:20.610 Samuel Roberts: The same thing in my.
382 00:33:20.610 ⇒ 00:33:21.070 Mustafa Raja: multiple places.
383 00:33:21.070 ⇒ 00:33:24.309 Samuel Roberts: but in the dock, it’s only in one place. Maybe that would help.
384 00:33:26.380 ⇒ 00:33:31.169 Samuel Roberts: That way, when we change things, it updates for everything, instead of we have to update it in 30 places.
385 00:33:31.690 ⇒ 00:33:32.500 Mustafa Raja: Oh, yeah, that…
386 00:33:32.500 ⇒ 00:33:33.210 Samuel Roberts: Do they have to update it.
387 00:33:33.210 ⇒ 00:33:33.740 Mustafa Raja: Yeah.
388 00:33:33.740 ⇒ 00:33:39.510 Samuel Roberts: Yeah, so if it’s like, this SOP applies to these 5 departments, when we rag it, we make sure to put it.
389 00:33:39.920 ⇒ 00:33:44.490 Mustafa Raja: Yeah, yeah, we could easily do that by assigning, IDs, too.
390 00:33:44.670 ⇒ 00:33:46.689 Mustafa Raja: Oh, she had sections, and then…
391 00:33:46.690 ⇒ 00:33:48.070 Samuel Roberts: That’s not a bad idea. Okay.
392 00:33:53.560 ⇒ 00:33:55.600 Amber Lin: Any other concerns?
393 00:34:09.300 ⇒ 00:34:13.850 Samuel Roberts: Not offhand right now, I think this covers a lot of the stuff I was worried about.
394 00:34:13.850 ⇒ 00:34:14.620 Amber Lin: Okay.
395 00:34:14.620 ⇒ 00:34:18.800 Samuel Roberts: I think the personnel is really the thing I’d rather keep out if best is possible, but I don’t know if it’s…
396 00:34:18.800 ⇒ 00:34:19.340 Amber Lin: C.
397 00:34:20.110 ⇒ 00:34:21.100 Samuel Roberts: Doable or not.
398 00:34:21.679 ⇒ 00:34:25.110 Amber Lin: So, what would be our next steps?
399 00:34:33.360 ⇒ 00:34:38.400 Samuel Roberts: Maybe we try to take… Which stock did you look at, Mustafa? The mechanical?
400 00:34:38.400 ⇒ 00:34:39.960 Mustafa Raja: Mechanical, yeah.
401 00:34:40.270 ⇒ 00:34:40.790 Samuel Roberts: Okay.
402 00:34:41.300 ⇒ 00:34:46.529 Amber Lin: I would say Tessa’s a better one that’s closer to this structure, mechanical.
403 00:34:46.530 ⇒ 00:34:47.500 Samuel Roberts: Yeah, yeah.
404 00:34:47.790 ⇒ 00:34:56.400 Samuel Roberts: Well, mechanical, we were trying to look at to compare it to pest, but if we just take… maybe we start with pest, and we just see, like, how close is it to this? How far off do we have to get to?
405 00:34:57.920 ⇒ 00:35:01.679 Amber Lin: Say, next steps, keep writing, I’ll write it now.
406 00:35:02.310 ⇒ 00:35:14.270 Mustafa Raja: Yeah, I was wondering if, Sam, if we are going to apply the same strategy we discussed, to see if that… that helped, select better sections or better pairs to compare.
407 00:35:15.830 ⇒ 00:35:20.809 Samuel Roberts: That, I’m not sure about for Pest.
408 00:35:23.590 ⇒ 00:35:25.709 Amber Lin: Wait, okay, I hear that we have.
409 00:35:25.710 ⇒ 00:35:26.310 Samuel Roberts: Ellen, yeah.
410 00:35:26.310 ⇒ 00:35:32.120 Amber Lin: We have some sort of process we’re doing, can we write that down?
411 00:35:32.120 ⇒ 00:35:36.440 Samuel Roberts: What was the duplicate and contradictory analysis that he was running?
412 00:35:36.760 ⇒ 00:35:42.850 Amber Lin: I see, I see. But do you think when we… when we’re trying to extract
413 00:35:42.930 ⇒ 00:35:53.130 Amber Lin: coverage for a certain service, we’ll still need to select chunks to compare. My goal here is to say, okay,
414 00:35:53.200 ⇒ 00:36:07.179 Amber Lin: how… how does an AI help us do this? What are the steps we need to go through so we can write it down and then look at the text instead of giving it? So, what would it look like? It would take a… take a…
415 00:36:07.290 ⇒ 00:36:11.030 Amber Lin: Take the central dock and chunk it first, or…
416 00:36:11.140 ⇒ 00:36:12.049 Amber Lin: How, how would that.
417 00:36:12.050 ⇒ 00:36:16.399 Mustafa Raja: No, on the go, if we want it on the go, I don’t think that…
418 00:36:17.300 ⇒ 00:36:19.789 Mustafa Raja: I don’t think that we would want to index it.
419 00:36:20.480 ⇒ 00:36:21.780 Samuel Roberts: No, I don’t think so.
420 00:36:22.400 ⇒ 00:36:24.640 Amber Lin: Do we need to index it ever?
421 00:36:24.760 ⇒ 00:36:26.850 Amber Lin: Like, would it be already indexed.
422 00:36:26.850 ⇒ 00:36:33.570 Mustafa Raja: Personally, to me, And correct me if I’m wrong, I don’t think we would want to…
423 00:36:33.700 ⇒ 00:36:37.190 Mustafa Raja: Index it… index it for this analysis at all, because…
424 00:36:38.350 ⇒ 00:36:41.109 Amber Lin: When you say this analysis, what are you referring to?
425 00:36:41.110 ⇒ 00:36:47.069 Mustafa Raja: for deduplication and, finding out the contradictions.
426 00:36:47.750 ⇒ 00:36:55.320 Amber Lin: I see, I see. I think I’m trying to get at, overall, If we were to establish…
427 00:36:55.500 ⇒ 00:36:58.760 Amber Lin: Sorry, sorry, let me gather my thoughts.
428 00:37:01.580 ⇒ 00:37:08.860 Amber Lin: I guess there’s one layer of overall, okay, we want to create something that’s more automatic, that improves the central dock.
429 00:37:09.030 ⇒ 00:37:22.769 Amber Lin: Under that, then we have different steps of, okay, we looked at what was the most asked questions, so that was the first step, and then we looked at, okay, we need to propose a structure and reorganize and rewrite the document.
430 00:37:23.150 ⇒ 00:37:30.790 Amber Lin: and then… I guess, how would an AI…
431 00:37:30.990 ⇒ 00:37:39.670 Amber Lin: do that? Is there any, like, technical steps we should write down? Because I need Drew guys’ input on that side, because I don’t know.
432 00:37:44.990 ⇒ 00:37:51.599 Mustafa Raja: If we’re trying to lag it, we… sorry, if we’re trying to, you know, chunk it and embed it, we already have that.
433 00:37:51.730 ⇒ 00:37:54.920 Mustafa Raja: So we don’t have to… Right, because you’re talking about, like.
434 00:37:54.920 ⇒ 00:37:58.640 Samuel Roberts: Restructuring it, like, what I need to do that.
435 00:37:59.190 ⇒ 00:38:04.909 Samuel Roberts: I mean, maybe what we do… I think this is probably it. We have to figure out if we… if these are the, like…
436 00:38:05.540 ⇒ 00:38:12.729 Samuel Roberts: big headlines, we probably want the AI to go through and figure out what needs to get moved to the right places.
437 00:38:12.890 ⇒ 00:38:21.689 Samuel Roberts: So… like, what is a shared SOP versus a specific SOP is something we could probably identify that way.
438 00:38:21.690 ⇒ 00:38:24.890 Mustafa Raja: I don’t know if we necessarily need to embed… Hmm.
439 00:38:24.890 ⇒ 00:38:26.760 Samuel Roberts: Probably just do that with the text.
440 00:38:27.280 ⇒ 00:38:28.010 Samuel Roberts: But…
441 00:38:34.200 ⇒ 00:38:38.920 Samuel Roberts: I don’t know if I have a… I don’t know if I have a better specific next steps than that, to be honest.
442 00:38:42.670 ⇒ 00:38:44.090 Samuel Roberts: So,
443 00:38:50.270 ⇒ 00:38:51.860 Samuel Roberts: Yeah, I would say…
444 00:38:52.390 ⇒ 00:38:57.090 Samuel Roberts: I’m not familiar enough with the Pestock to know how close it is to this structure, but…
445 00:38:58.280 ⇒ 00:39:00.819 Samuel Roberts: If we basically want to take…
446 00:39:01.520 ⇒ 00:39:04.409 Samuel Roberts: Something like, say, the pest dock, because it’s closest.
447 00:39:04.560 ⇒ 00:39:13.520 Samuel Roberts: Pass that to… Opus with a new proposed structure, and ask it
448 00:39:14.090 ⇒ 00:39:17.569 Samuel Roberts: what needs to go where.
449 00:39:18.220 ⇒ 00:39:21.110 Samuel Roberts: The context window’s probably big enough for all that.
450 00:39:25.090 ⇒ 00:39:28.340 Samuel Roberts: I don’t know, does that… Seemed like a…
451 00:39:28.520 ⇒ 00:39:37.260 Amber Lin: Would we also want to run the path of, is this the best structure? Like, we can use the paths as an example, but…
452 00:39:37.590 ⇒ 00:39:44.580 Samuel Roberts: Yeah, I mean, that’s a good… yeah. No, I think, yeah, I would say, like, you’re right, this is a proposed structure, not necessarily the final one, so I would say, yeah, that’s probably another…
453 00:39:45.190 ⇒ 00:39:48.019 Samuel Roberts: another pass to do there. Like, given all this information.
454 00:39:48.020 ⇒ 00:39:48.770 Amber Lin: subject.
455 00:39:48.770 ⇒ 00:39:55.150 Samuel Roberts: the structure, we think this is not bad, but it’s not started. It’s not the finalized thing, yeah.
456 00:39:58.290 ⇒ 00:40:00.360 Amber Lin: Yeah, and then we can also…
457 00:40:01.140 ⇒ 00:40:13.460 Amber Lin: consider the type of questions asked to also suggest, areas of information that people want, but we might not always have. So we can use, like, a test.
458 00:40:13.700 ⇒ 00:40:14.859 Amber Lin: Central thought…
459 00:40:14.860 ⇒ 00:40:15.570 Samuel Roberts: Right, right.
460 00:40:15.570 ⇒ 00:40:18.359 Amber Lin: And some questions.
461 00:40:18.890 ⇒ 00:40:20.280 Amber Lin: Response.
462 00:40:21.570 ⇒ 00:40:22.580 Amber Lin: Okay.
463 00:40:26.010 ⇒ 00:40:38.240 Amber Lin: Because I think the structure we arrived at with… with Pranav’s analysis of this is the areas people were asking, is helpful for our overall structure.
464 00:40:38.650 ⇒ 00:40:39.900 Amber Lin: as well.
465 00:40:39.900 ⇒ 00:40:41.290 Samuel Roberts: Yes, definitely.
466 00:40:41.510 ⇒ 00:40:53.130 Amber Lin: Cool, okay. So once we do that, I guess we then look at what goes where, or at least tag them with each category in this structure.
467 00:40:53.250 ⇒ 00:40:53.940 Mustafa Raja: Hmm.
468 00:40:54.710 ⇒ 00:40:57.000 Amber Lin: Will we do that, or… Yeah. Okay, so…
469 00:40:57.000 ⇒ 00:41:01.109 Samuel Roberts: Yeah, I think we could probably even get it to reflow the document, potentially.
470 00:41:02.410 ⇒ 00:41:02.920 Amber Lin: Okay.
471 00:41:02.920 ⇒ 00:41:09.840 Samuel Roberts: its own, I don’t know. I have to look at the actual length of the stuff, but I bet it could probably just give us a good pass.
472 00:41:11.530 ⇒ 00:41:21.890 Amber Lin: Text with section, it should… belong, and then I guess we can do a manual, or AI pass?
473 00:41:21.890 ⇒ 00:41:23.609 Samuel Roberts: Yeah, either way.
474 00:41:23.860 ⇒ 00:41:28.090 Amber Lin: of… It’s very long.
475 00:41:29.180 ⇒ 00:41:37.980 Amber Lin: Or not. And then… Lastly, AI… Temp and restructuring…
476 00:41:39.210 ⇒ 00:41:42.669 Amber Lin: I mean, at one of these points.
477 00:41:42.950 ⇒ 00:42:02.039 Amber Lin: we would want the AI to tell us, hey, this is contradictory, this is wrong, this might be, a rewrite of a very similar thing, would you like to combine it, right? Like, we would want some checkpoints where we can go to the trainers, like, hey, this is where we need your input. Where would that be?
478 00:42:04.650 ⇒ 00:42:11.509 Samuel Roberts: I mean, if we’re tagging existing text with sections it would belong in, that might be, like, we might see several where…
479 00:42:11.800 ⇒ 00:42:14.910 Amber Lin: Oh, these all got tagged, and they look very similar.
480 00:42:16.460 ⇒ 00:42:21.160 Samuel Roberts: the AI might be able to highlight that, and we might even be able to highlight that, depending on how obvious it is.
481 00:42:21.670 ⇒ 00:42:25.009 Amber Lin: Okay, so we’ll say this is the step where we…
482 00:42:25.390 ⇒ 00:42:25.840 Samuel Roberts: Yeah.
483 00:42:25.840 ⇒ 00:42:31.359 Amber Lin: Confirm and gather information from trainers.
484 00:42:32.810 ⇒ 00:42:41.580 Amber Lin: To combine… Rewrite… So true. Okay.
485 00:42:42.470 ⇒ 00:42:47.080 Amber Lin: Input… Alright.
486 00:42:47.660 ⇒ 00:42:52.529 Amber Lin: Let’s say… I’ll put a…
487 00:42:53.810 ⇒ 00:43:03.169 Amber Lin: when we want… let’s talk about what the deliverable is for each step, then. So I think we’ll make the execution a bit clearer.
488 00:43:03.580 ⇒ 00:43:20.730 Amber Lin: What do you need for a structure? Do you need, like, metadata tags? Do you need a description of what it is, or just sample data? Or do you even want, like, markdown structure of this goes first, that goes where? This is, like…
489 00:43:21.010 ⇒ 00:43:26.419 Amber Lin: like, formatted this way. How… how would you define a structure?
490 00:43:27.230 ⇒ 00:43:29.130 Samuel Roberts: Yeah, I would think it would be, like, a…
491 00:43:30.730 ⇒ 00:43:35.099 Samuel Roberts: Hierarchy of headings, effectively, with, like, maybe…
492 00:43:36.290 ⇒ 00:43:44.489 Samuel Roberts: You know, if we’re keeping the structure and not any content, maybe a little bit of description about what is in there, what should go in those sections.
493 00:43:46.460 ⇒ 00:43:57.409 Samuel Roberts: but not actually the content itself. So, it’d probably be, like, you know, headings, like, you know, SOP, and then, like, scheduling. Like, it might be, like, a few of those things that are, like, big ones, but…
494 00:43:58.540 ⇒ 00:44:02.269 Samuel Roberts: not… The content of those.
495 00:44:03.290 ⇒ 00:44:08.429 Amber Lin: Okay, then how would the AI know when we use it
496 00:44:08.580 ⇒ 00:44:12.740 Amber Lin: To, like, tag things or classify things, like, how would it know
497 00:44:12.980 ⇒ 00:44:20.140 Amber Lin: where it belongs? Would it look at how the text is generally structured, or would it just look for keywords?
498 00:44:21.100 ⇒ 00:44:26.230 Samuel Roberts: I think you would have to look at the content and how it’s currently organized, and try to infer it from that.
499 00:44:29.040 ⇒ 00:44:29.850 Amber Lin: Okay.
500 00:44:30.210 ⇒ 00:44:33.540 Samuel Roberts: Because there’s some structure, it’s just not the structure that we think is the most beneficial.
501 00:44:33.540 ⇒ 00:44:34.010 Amber Lin: Yeah.
502 00:44:34.010 ⇒ 00:44:34.540 Samuel Roberts: You know.
503 00:44:34.700 ⇒ 00:44:35.859 Amber Lin: Yeah, okay.
504 00:44:36.590 ⇒ 00:44:47.210 Amber Lin: I guess, Casey and Mustafa, you guys would be executing these, like, what other… Imagine if this was a ticket, like, what else information would you…
505 00:44:56.810 ⇒ 00:44:57.540 Casie Aviles: Hello.
506 00:44:57.760 ⇒ 00:44:59.140 Samuel Roberts: What do you guys think? Yeah.
507 00:45:01.100 ⇒ 00:45:04.720 Casie Aviles: Sorry, which, which part are we talking about specifically?
508 00:45:05.290 ⇒ 00:45:09.009 Amber Lin: In the next steps, we can look at the step one.
509 00:45:09.360 ⇒ 00:45:19.879 Amber Lin: I guess these 3 or 4 steps of… like, if it was a ticket, what info do you need to, like, clearly execute?
510 00:45:20.030 ⇒ 00:45:21.160 Amber Lin: the task.
511 00:45:22.250 ⇒ 00:45:23.719 Amber Lin: Since we’re all here.
512 00:45:24.910 ⇒ 00:45:27.359 Amber Lin: Because I know that it’s kind of unclear right now.
513 00:45:28.880 ⇒ 00:45:32.649 Casie Aviles: Yeah Don’t you understand?
514 00:45:33.110 ⇒ 00:45:41.140 Casie Aviles: Well, confirm this, I’m sure… Well, if we want to…
515 00:45:41.260 ⇒ 00:45:44.700 Casie Aviles: Confirm the best structure right now, then.
516 00:45:46.200 ⇒ 00:45:52.899 Casie Aviles: I think… There’s going to… it’s going to involve some testing with Andy.
517 00:45:54.360 ⇒ 00:46:04.280 Casie Aviles: Hmm. I… because we want to be able to, like, know if the responses will also…
518 00:46:05.270 ⇒ 00:46:07.169 Casie Aviles: You know, are better.
519 00:46:09.660 ⇒ 00:46:12.990 Casie Aviles: I’m not sure if I… if I am… I have the right click.
520 00:46:12.990 ⇒ 00:46:16.970 Samuel Roberts: Oh, I see what you’re saying, yeah. Like, how do we know that it’s… I’m thinking…
521 00:46:17.850 ⇒ 00:46:28.610 Samuel Roberts: Just even if we’re just, you know, condensing and deduplicating, that will be… That’ll be enough of a…
522 00:46:29.010 ⇒ 00:46:35.590 Samuel Roberts: Gain, that we can probably put off a little bit of testing until we have more of a reflow document.
523 00:46:37.330 ⇒ 00:46:40.029 Samuel Roberts: Sample questions are good, yeah.
524 00:46:49.240 ⇒ 00:46:56.479 Samuel Roberts: I mean, this might even be something we just need to give to Opus and see, like, knowing… making sure that it knows what we’re trying to do with it, too.
525 00:46:57.700 ⇒ 00:46:58.780 Amber Lin: Okay.
526 00:46:59.240 ⇒ 00:47:01.130 Casie Aviles: So, a system prompt then.
527 00:47:02.350 ⇒ 00:47:04.020 Samuel Roberts: Excuse me, sorry.
528 00:47:08.300 ⇒ 00:47:17.920 Amber Lin: I mean, if we really don’t know much on the requirements, we can just give it a rough spike timeline, and then we can talk about it once we do a discovery.
529 00:47:18.880 ⇒ 00:47:24.059 Samuel Roberts: probably smart, because I think there will be some back-and-forth iterations that we’re going to want to do on this, so…
530 00:47:24.540 ⇒ 00:47:40.189 Amber Lin: Cool. Well, then… then my main ask is, can I have a milestone deadline, so that the client can know, like, oh, we should wait this… this many weeks, and this is when…
531 00:47:40.430 ⇒ 00:47:43.860 Amber Lin: they’ll need our input. Like, I’m meeting with them, I think, in…
532 00:47:44.500 ⇒ 00:47:58.240 Amber Lin: two weeks? So, next Friday. So, not this Friday, next Friday, I’m meeting with them again. What state will we be at? Like, what can I show them? What can I tell them? What can I ask them about? Will we be.
533 00:47:58.240 ⇒ 00:47:59.040 Samuel Roberts: Yeah.
534 00:47:59.040 ⇒ 00:48:00.979 Amber Lin: stage, when I meet with them?
535 00:48:00.980 ⇒ 00:48:05.589 Samuel Roberts: I think we could probably get to a point when we would have a,
536 00:48:07.430 ⇒ 00:48:11.850 Samuel Roberts: An ideal structure, maybe even put, like, pest into that structure.
537 00:48:12.750 ⇒ 00:48:13.210 Samuel Roberts: Mmm.
538 00:48:14.400 ⇒ 00:48:15.140 Casie Aviles: Yeah, I think we won’.
539 00:48:15.140 ⇒ 00:48:18.679 Samuel Roberts: And then say, like, this is how we’re gonna try to match the… sorry?
540 00:48:19.670 ⇒ 00:48:24.090 Casie Aviles: Yeah, yeah, I was just saying, like, we want to confirm, like, the best structure first, right?
541 00:48:24.090 ⇒ 00:48:33.099 Samuel Roberts: Yeah, so I think if we can do that, we may even be able to, like, point Andy to it and test it by then, I don’t know if that’s feasible.
542 00:48:33.260 ⇒ 00:48:41.959 Mustafa Raja: So we are going to create this structure, and then we are going to point Andy to see if it answers good. Is that correct?
543 00:48:42.770 ⇒ 00:48:45.669 Casie Aviles: Yeah, that’s kind of what I was thinking, where we…
544 00:48:45.670 ⇒ 00:48:47.449 Samuel Roberts: Yeah, I think we’re gonna have to do that eventually.
545 00:48:47.630 ⇒ 00:48:52.780 Casie Aviles: What we have right now, versus, you know, what the… New structure is.
546 00:48:53.120 ⇒ 00:48:59.060 Amber Lin: Oh, for Andy to test it, don’t we have to have AI rewrite the document already?
547 00:48:59.060 ⇒ 00:49:00.130 Mustafa Raja: Yeah…
548 00:49:00.730 ⇒ 00:49:05.350 Samuel Roberts: Also, I think we can… I don’t know if we’ll get that far by next Friday, but… Yeah.
549 00:49:05.350 ⇒ 00:49:05.740 Casie Aviles: I see.
550 00:49:05.740 ⇒ 00:49:06.639 Amber Lin: I do think we…
551 00:49:06.640 ⇒ 00:49:07.540 Samuel Roberts: Hopefully we can get to it.
552 00:49:07.540 ⇒ 00:49:12.549 Amber Lin: testing, but, like, I think this is good, like, using passive as an example.
553 00:49:12.800 ⇒ 00:49:20.040 Samuel Roberts: Yeah, I think if we can get that, like, you know, learn the structure, see what it… see how we can improve it, like, kind of be like, here’s the…
554 00:49:20.390 ⇒ 00:49:22.229 Samuel Roberts: You know, before and after.
555 00:49:22.830 ⇒ 00:49:27.020 Samuel Roberts: And let them know that we’re gonna… test that from there. Cool.
556 00:49:27.120 ⇒ 00:49:30.579 Amber Lin: Okay, so I’ll say this is Violet next.
557 00:49:31.340 ⇒ 00:49:39.150 Amber Lin: Friday… And this one by this Friday? Is that good? Is that enough time?
558 00:49:40.170 ⇒ 00:49:41.500 Samuel Roberts: Which one? Sorry about that.
559 00:49:41.500 ⇒ 00:49:43.390 Amber Lin: So, the step one, confirming vaccine.
560 00:49:43.390 ⇒ 00:49:45.510 Samuel Roberts: Yeah, I think so, I think so.
561 00:49:45.990 ⇒ 00:49:50.519 Amber Lin: So I’ll say Friday. Cool. Yeah, that’s all I need.
562 00:49:51.040 ⇒ 00:49:54.300 Mustafa Raja: So, so this step one does include,
563 00:49:54.480 ⇒ 00:49:59.149 Mustafa Raja: anti-testing it, right? So, so we using Office, creating a…
564 00:49:59.150 ⇒ 00:49:59.729 Samuel Roberts: I don’t know.
565 00:49:59.770 ⇒ 00:50:01.260 Mustafa Raja: So…
566 00:50:02.860 ⇒ 00:50:12.660 Amber Lin: Let me put here, Andy testing would… Re… Rewritten box.
567 00:50:14.180 ⇒ 00:50:17.640 Amber Lin: How is this? We probably need to do it on mechanical.
568 00:50:17.990 ⇒ 00:50:20.449 Amber Lin: The mechanical has the most issues.
569 00:50:21.110 ⇒ 00:50:21.670 Samuel Roberts: Okay.
570 00:50:22.810 ⇒ 00:50:28.419 Casie Aviles: So what I’m thinking for the first one is we’re just going to use the best one as…
571 00:50:29.450 ⇒ 00:50:30.000 Amber Lin: Yeah.
572 00:50:30.000 ⇒ 00:50:30.910 Casie Aviles: Golden standard.
573 00:50:31.590 ⇒ 00:50:35.610 Amber Lin: I think so, and then we’re gonna take that to confront the trainers.
574 00:50:36.480 ⇒ 00:50:37.010 Mustafa Raja: Okay.
575 00:50:37.010 ⇒ 00:50:39.760 Casie Aviles: And then we’ll probably pass it through…
576 00:50:39.870 ⇒ 00:50:44.770 Casie Aviles: AI, I guess, to also, like, have any… get any recommendations.
577 00:50:46.590 ⇒ 00:50:48.759 Casie Aviles: Okay. Yeah, those are kind of, like.
578 00:50:49.110 ⇒ 00:50:51.889 Casie Aviles: Oh, I’m thinking we could go about this.
579 00:50:52.340 ⇒ 00:50:53.360 Casie Aviles: First step.
580 00:50:53.910 ⇒ 00:50:55.530 Amber Lin: Okay, okay.
581 00:50:55.720 ⇒ 00:50:56.840 Amber Lin: Sounds good.
582 00:50:57.280 ⇒ 00:51:00.149 Amber Lin: We feel more confident about the timeline.
583 00:51:02.240 ⇒ 00:51:09.090 Casie Aviles: Yeah, I think by Friday, that should do… I’m not sure if I’ll be taking that pill, but… Yep.
584 00:51:09.090 ⇒ 00:51:12.310 Amber Lin: Yeah, so would you have time for that?
585 00:51:16.990 ⇒ 00:51:18.430 Amber Lin: Because we… I think we can pause…
586 00:51:18.430 ⇒ 00:51:21.090 Mustafa Raja: So, we’re done with this one, right?
587 00:51:23.970 ⇒ 00:51:38.530 Amber Lin: Yeah, I’m just gonna mark it as done for now. I think we found enough stuff that interested the clients. They were like, oh, that’s cool, and they had more buy-in. So right now, I think we can move on to, like, overall structure.
588 00:51:38.530 ⇒ 00:51:40.390 Mustafa Raja: Yeah, I can take this one, because I’m…
589 00:51:40.390 ⇒ 00:51:41.150 Amber Lin: soon.
590 00:51:41.620 ⇒ 00:51:46.270 Mustafa Raja: Yeah, yeah, yeah, I was going to work on this one, so I’m going to work on this one then.
591 00:51:47.930 ⇒ 00:51:55.860 Amber Lin: Sure, okay, cool, sounds good. Let’s say… I’m gonna assign this to you…
592 00:51:59.410 ⇒ 00:52:00.420 Amber Lin: Okay.
593 00:52:09.960 ⇒ 00:52:11.190 Amber Lin: Alright.
594 00:52:11.950 ⇒ 00:52:19.590 Amber Lin: Okay, we have, like, 9 minutes left. Sam, what is our next steps on the transcripts?
595 00:52:20.500 ⇒ 00:52:25.280 Samuel Roberts: Yeah, Utam wanted me to get just the sample into BigQuery.
596 00:52:26.920 ⇒ 00:52:34.730 Samuel Roberts: So that you guys could start doing some stuff with it, I guess. I didn’t… it’s not really built into full transcripts, but all the data’s in there.
597 00:52:35.860 ⇒ 00:52:37.139 Samuel Roberts: I’m not really sure.
598 00:52:37.840 ⇒ 00:52:44.530 Samuel Roberts: what that is next, I… I don’t know who is taking that, kinda, but I have…
599 00:52:44.530 ⇒ 00:52:45.930 Amber Lin: Probably be me.
600 00:52:46.250 ⇒ 00:52:47.810 Amber Lin: Okay. I have…
601 00:52:47.810 ⇒ 00:52:49.999 Samuel Roberts: In which case, yeah, I don’t know…
602 00:52:50.130 ⇒ 00:52:58.120 Samuel Roberts: I can try to help out with BigQuery. I don’t know a ton about BigQuery necessarily, but I know a little more about these transcripts now, so I don’t know…
603 00:52:58.820 ⇒ 00:52:59.400 Amber Lin: I see.
604 00:53:00.380 ⇒ 00:53:01.240 Amber Lin: Okay.
605 00:53:01.240 ⇒ 00:53:06.040 Samuel Roberts: you need most. Like, all the data’s in there, I’m not sure… how…
606 00:53:06.910 ⇒ 00:53:11.490 Samuel Roberts: how you need to work with it best, so I… I kind of defer to you on that, but I have…
607 00:53:12.200 ⇒ 00:53:17.000 Samuel Roberts: downloaded the metadata for the next, like, the full month of January.
608 00:53:17.240 ⇒ 00:53:17.640 Amber Lin: Okay.
609 00:53:17.640 ⇒ 00:53:22.110 Samuel Roberts: So… We could start loading that at some point, it will take a little bit, but…
610 00:53:22.530 ⇒ 00:53:23.210 Amber Lin: Okay.
611 00:53:23.840 ⇒ 00:53:27.190 Samuel Roberts: Take a look at what’s in there, let me know what questions you have.
612 00:53:27.550 ⇒ 00:53:28.200 Amber Lin: Nope.
613 00:53:28.200 ⇒ 00:53:28.760 Samuel Roberts: Yeah.
614 00:53:29.130 ⇒ 00:53:30.760 Amber Lin: Yeah,
615 00:53:31.500 ⇒ 00:53:43.300 Amber Lin: I know we were also trying to do, like, the Andy and transcript response parry. Are we still doing that, or are you blocked on that?
616 00:53:44.370 ⇒ 00:53:46.080 Samuel Roberts: I… the pairing?
617 00:53:46.590 ⇒ 00:53:53.980 Amber Lin: So, like, what call… Was… is related to what message?
618 00:53:54.300 ⇒ 00:54:05.049 Samuel Roberts: Oh… I… Don’t… Know the best way to handle that yet.
619 00:54:05.290 ⇒ 00:54:07.560 Samuel Roberts: Based on what I’ve seen in the transcripts.
620 00:54:07.780 ⇒ 00:54:08.270 Amber Lin: I see.
621 00:54:08.940 ⇒ 00:54:16.209 Samuel Roberts: There may be some things we can do with correlating time, and seeing what was asked, but I… that… I’m not sure, I haven’t dug into that enough.
622 00:54:16.830 ⇒ 00:54:21.369 Amber Lin: I see, so that would be probably, like, time and then person.
623 00:54:22.030 ⇒ 00:54:23.060 Samuel Roberts: Yeah…
624 00:54:24.850 ⇒ 00:54:29.089 Amber Lin: I see, because they don’t take two calls at the same time, so it… Right.
625 00:54:29.090 ⇒ 00:54:36.690 Samuel Roberts: Right, so yeah, one person… that’s true, we should be able to tie it to the person, which I think I do have access… the data is there now. At first, I didn’t think I had access to it, but I found that.
626 00:54:36.690 ⇒ 00:54:37.680 Amber Lin: Okay.
627 00:54:37.870 ⇒ 00:54:46.490 Amber Lin: So, I will do my initial… Analysis, discovery… Pairing, and then we’ll do…
628 00:54:49.780 ⇒ 00:54:52.329 Amber Lin: I’m gonna delete this one.
629 00:54:52.740 ⇒ 00:54:56.420 Amber Lin: Cool, okay, so I’ll… I’ll go ahead and take a look there.
630 00:54:56.960 ⇒ 00:54:58.819 Samuel Roberts: Okay, yeah, let me know,
631 00:54:59.380 ⇒ 00:55:04.460 Samuel Roberts: Because, like, yeah, they’re… and they’re doubled up because there’s two channels per call for the…
632 00:55:04.790 ⇒ 00:55:13.679 Samuel Roberts: caller and the agent, and so you have to kind of join them together, so I’m happy to, like, help out there, but I just dumped it all, because that’s what Utam wanted, to just get it all.
633 00:55:13.680 ⇒ 00:55:14.090 Amber Lin: Understood.
634 00:55:14.090 ⇒ 00:55:15.170 Samuel Roberts: for now, so…
635 00:55:15.170 ⇒ 00:55:19.200 Amber Lin: I can do the initial analysis if it, if it needs, like.
636 00:55:19.910 ⇒ 00:55:23.150 Amber Lin: Not modeling, but, say, formatting changes or extracting.
637 00:55:23.150 ⇒ 00:55:23.680 Samuel Roberts: Yeah.
638 00:55:23.680 ⇒ 00:55:27.010 Amber Lin: Whatever, I will ask for help, because I have no clue how to do that.
639 00:55:27.010 ⇒ 00:55:28.730 Samuel Roberts: Okay, cool, yeah, sounds good.
640 00:55:29.110 ⇒ 00:55:30.610 Amber Lin: Okay, yeah, that’s all.
641 00:55:30.810 ⇒ 00:55:33.670 Amber Lin: Thanks, everybody. I’ll make sure the teams are…
642 00:55:33.810 ⇒ 00:55:48.980 Mustafa Raja: Yeah. Yeah, just to confirm the step one, what’s going to happen there is we are going to put in, test doc, and then the questions from, Pranav, and then have Opus go through them, suggest a structure. Is that correct?
643 00:55:48.980 ⇒ 00:55:58.719 Amber Lin: Yeah, and if FNAF doesn’t have time to give it to you, you can just use the raw data. I think you know where it is, so just… just dump that in, and…
644 00:55:58.720 ⇒ 00:56:03.120 Mustafa Raja: Okay, so it’s the same data that’s in Snowflake, right?
645 00:56:03.280 ⇒ 00:56:04.239 Amber Lin: Yeah, that one.
646 00:56:04.240 ⇒ 00:56:05.440 Samuel Roberts: Okay, fantastic.
647 00:56:05.980 ⇒ 00:56:06.520 Amber Lin: Cool.
648 00:56:06.520 ⇒ 00:56:07.400 Samuel Roberts: Okay.
649 00:56:07.850 ⇒ 00:56:08.700 Mustafa Raja: Thank you.
650 00:56:09.150 ⇒ 00:56:09.700 Samuel Roberts: Yeah.
651 00:56:10.430 ⇒ 00:56:11.130 Amber Lin: Alright, bye-bye.
652 00:56:11.650 ⇒ 00:56:12.470 Samuel Roberts: Right.