Meeting Title: ANDI Project Weekly Planning Date: 2026-03-10 Meeting participants: Pranav Narahari, Casie Aviles, Samuel Roberts, Amber Lin, Mustafa Raja
WEBVTT
1 00:02:44.110 ⇒ 00:02:45.070 Samuel Roberts: Hello.
2 00:02:48.350 ⇒ 00:02:49.030 Pranav Narahari: Hey, guys.
3 00:02:50.300 ⇒ 00:02:51.020 Amber Lin: Hello.
4 00:02:51.020 ⇒ 00:02:51.780 Pranav Narahari: Hello.
5 00:02:53.450 ⇒ 00:02:55.640 Pranav Narahari: Hey, Amber. Hey, Sam. Hey, Casey.
6 00:02:57.590 ⇒ 00:02:58.430 Samuel Roberts: Interesting.
7 00:03:06.350 ⇒ 00:03:07.270 Samuel Roberts: Okay, there we go.
8 00:03:07.470 ⇒ 00:03:09.979 Pranav Narahari: Anybody know if Mustafa’s joining?
9 00:03:14.260 ⇒ 00:03:15.450 Samuel Roberts: I can ping him.
10 00:03:16.300 ⇒ 00:03:17.530 Pranav Narahari: Okay, perfect.
11 00:03:48.070 ⇒ 00:03:50.280 Pranav Narahari: Perfect. Hey, Mustafa.
12 00:03:54.010 ⇒ 00:04:02.699 Pranav Narahari: Okay, let’s, let’s just get started. There’s gonna be a bunch of different colors here. I’ll kind of explain, like, what I did to the Gantt chart.
13 00:04:02.880 ⇒ 00:04:07.040 Pranav Narahari: So…
14 00:04:07.320 ⇒ 00:04:14.370 Pranav Narahari: there are a couple things, and I’ll kind of, like, run through, like, what I feel like are most urgent.
15 00:04:15.190 ⇒ 00:04:24.969 Pranav Narahari: this thing right here, the migrate thumbs up, thumbs down, I think this should be switched to Mustafa, but honestly, I’ll wait on that, too.
16 00:04:25.520 ⇒ 00:04:27.809 Pranav Narahari: What is the current status on this one?
17 00:04:28.110 ⇒ 00:04:28.900 Pranav Narahari: And, like, how close…
18 00:04:28.900 ⇒ 00:04:31.639 Mustafa Raja: I still need to start it up.
19 00:04:32.050 ⇒ 00:04:33.370 Mustafa Raja: Start migrating it.
20 00:04:33.840 ⇒ 00:04:35.180 Pranav Narahari: This needs to be started.
21 00:04:36.160 ⇒ 00:04:36.890 Pranav Narahari: Okay.
22 00:04:36.890 ⇒ 00:04:41.459 Samuel Roberts: Yeah, I think, if I recall correctly, that’s, like, kind of tied into the Google Chat
23 00:04:41.790 ⇒ 00:04:44.440 Samuel Roberts: code, as well as the monster code, right?
24 00:04:44.930 ⇒ 00:04:45.850 Mustafa Raja: Yes.
25 00:04:46.170 ⇒ 00:04:49.369 Samuel Roberts: So it wasn’t something we could do before we had all the Google stuff set up.
26 00:04:50.390 ⇒ 00:04:51.050 Pranav Narahari: Okay.
27 00:04:51.470 ⇒ 00:04:56.050 Pranav Narahari: So, yeah, we’re not waiting on anything now to get that going, right? So… I would say
28 00:04:56.890 ⇒ 00:05:00.390 Pranav Narahari: Okay, perfect. Just making sure there’s no blocker here.
29 00:05:02.570 ⇒ 00:05:08.550 Pranav Narahari: Let’s just kind of go over all the… Priorities, and then…
30 00:05:08.740 ⇒ 00:05:21.010 Pranav Narahari: we can talk about, just like, okay, how are we gonna tackle all of them? Yeah, so I’m putting this color just for the things that I think we can keep, that we should keep, I should say. This, like, this lighter red is,
31 00:05:21.610 ⇒ 00:05:27.929 Pranav Narahari: something where I kind of want to discuss more, can we push it down the line? Does it need to be prioritized right now?
32 00:05:28.710 ⇒ 00:05:34.000 Pranav Narahari: staging, validation, end-to-end, I believe we should have put this in completed, right?
33 00:05:35.470 ⇒ 00:05:41.390 Casie Aviles: This is in progress. Yeah, we’re doing this right now. That’s why we have all the testing.
34 00:05:43.780 ⇒ 00:05:44.660 Pranav Narahari: Okay.
35 00:05:44.730 ⇒ 00:05:49.119 Casie Aviles: The established one is the one that we put in completed, this one.
36 00:05:49.480 ⇒ 00:05:52.940 Pranav Narahari: I see, I see. Okay. Gotcha. And so…
37 00:05:54.040 ⇒ 00:06:00.950 Pranav Narahari: When we say validation, is that… is that, like, running through the triage tickets?
38 00:06:01.120 ⇒ 00:06:02.800 Pranav Narahari: What are we doing to validate it?
39 00:06:04.340 ⇒ 00:06:06.599 Casie Aviles: Yeah, for this one, it’s…
40 00:06:06.740 ⇒ 00:06:11.889 Casie Aviles: We’re not… I’m not really testing this on the triage. What I’m doing right now is…
41 00:06:12.110 ⇒ 00:06:16.030 Casie Aviles: Just running all of it through, like, all of the…
42 00:06:16.280 ⇒ 00:06:22.010 Casie Aviles: Snowflake logs, not necessarily just the triages or the ones that have thumbs down.
43 00:06:22.340 ⇒ 00:06:27.559 Casie Aviles: So, I’m… you know, I’m comparing, like, the execution times.
44 00:06:28.020 ⇒ 00:06:34.880 Casie Aviles: Across, like, the N810 ones and the ones we have for Astra that, you know, right now, the…
45 00:06:34.990 ⇒ 00:06:36.189 Casie Aviles: Setup we have.
46 00:06:38.870 ⇒ 00:06:47.620 Pranav Narahari: Okay, and are you just looking to see that it’s not… there’s no errors, or are you looking to just, like, compare the accuracy and the execution time?
47 00:06:49.510 ⇒ 00:06:56.540 Casie Aviles: Yes, that should be… that’s the idea. Right now, I’m just focusing on the execution times, but…
48 00:06:57.120 ⇒ 00:07:03.960 Casie Aviles: I’m also taking a look at whether… basically, it’s just a surface-level assessment of, like.
49 00:07:04.540 ⇒ 00:07:08.630 Casie Aviles: The accuracy, whether… whether it… answers.
50 00:07:08.780 ⇒ 00:07:14.889 Casie Aviles: Something… Or if it doesn’t. So that’s… that’s how I take a look at it right now.
51 00:07:17.860 ⇒ 00:07:18.780 Pranav Narahari: Gotcha.
52 00:07:20.210 ⇒ 00:07:23.950 Pranav Narahari: Okay, so there might be some duplicates here, is,
53 00:07:26.490 ⇒ 00:07:30.969 Pranav Narahari: But we’ll get down, we’ll get there. So, migrate AI Models, I believe that’s completed, right?
54 00:07:33.880 ⇒ 00:07:40.510 Casie Aviles: Yeah, that should be completed. We already have the Google Git, and I… Right. Models.
55 00:07:44.120 ⇒ 00:07:45.240 Pranav Narahari: Yep. Okay, cool.
56 00:07:45.590 ⇒ 00:07:54.269 Pranav Narahari: And so, does this milestone still make sense? Providing anti-staging access to the team.
57 00:07:54.640 ⇒ 00:08:02.099 Pranav Narahari: One, actually, question that I have here is, won’t this also be dependent on embedding the new central docs?
58 00:08:07.900 ⇒ 00:08:10.470 Samuel Roberts: It could be, yeah, I mean…
59 00:08:11.910 ⇒ 00:08:18.999 Amber Lin: I was thinking that could be separate, because it’s not completely dependent, like, this old Central Docs would still work.
60 00:08:21.230 ⇒ 00:08:22.320 Amber Lin: I…
61 00:08:22.520 ⇒ 00:08:34.659 Amber Lin: I’m not exactly sure when we decided on the migration, so I think our main improvements should be, like, answer accuracy and processing time, but…
62 00:08:36.380 ⇒ 00:08:39.870 Amber Lin: Like, in my impression, the central doc was a separate workflow.
63 00:08:41.679 ⇒ 00:08:45.849 Pranav Narahari: I believe Central Dock is what’s going to improve accuracy, right?
64 00:08:45.850 ⇒ 00:08:52.020 Amber Lin: Let’s see… I see. I wasn’t involved in when we started thinking about the migration.
65 00:08:52.020 ⇒ 00:08:52.910 Pranav Narahari: So…
66 00:08:52.950 ⇒ 00:08:59.420 Amber Lin: I’m actually not very certain what our goal was when we started the migration.
67 00:08:59.680 ⇒ 00:09:04.079 Pranav Narahari: Gotcha. Okay, so I’m pretty… yeah, the migration sounds like it was a fully, like…
68 00:09:04.260 ⇒ 00:09:07.919 Pranav Narahari: accuracy play, right? We’re trying to improve accuracy by having…
69 00:09:07.920 ⇒ 00:09:09.970 Casie Aviles: accuracy and latency.
70 00:09:10.480 ⇒ 00:09:15.199 Pranav Narahari: Okay, and execution time, it’s going to assist with that. Okay, so…
71 00:09:15.400 ⇒ 00:09:26.529 Pranav Narahari: I do like that staged approach, especially since, like, it requires less from us right now to deliver something to them. I want to deliver as soon as possible, I guess, on this. And…
72 00:09:26.670 ⇒ 00:09:30.350 Pranav Narahari: Specifically since, like, the central dock still needs some work, let’s…
73 00:09:30.980 ⇒ 00:09:35.940 Pranav Narahari: Let’s, like, stick to this milestone, meaning that, like, it’s not using the restructured
74 00:09:36.230 ⇒ 00:09:48.370 Pranav Narahari: central dock, which can… is going to help further reduce execution time and accuracy. Providing Andy’s staging
75 00:09:48.660 ⇒ 00:10:04.230 Pranav Narahari: I haven’t said anything specifically to them about, like, we’re gonna give them staging, so we could even push this milestone down a little bit, because what is the main priority is going to be getting those central docs to them by tomorrow end of day.
76 00:10:05.590 ⇒ 00:10:11.749 Pranav Narahari: So… Yeah, we can… I mean, I guess Amber, like…
77 00:10:11.930 ⇒ 00:10:15.339 Pranav Narahari: maybe I can just use a little bit of extra context here, like…
78 00:10:15.340 ⇒ 00:10:32.400 Amber Lin: Yeah, I think when we say provide access to team, we essentially just mean, Janice and Yvette, or it may be the department leads. So, this is not going to be their full team, this is mostly to show, hey, test this out, this is our new version, it’s better.
79 00:10:32.400 ⇒ 00:10:42.009 Amber Lin: get a feel for it, we’re still kind of developing, so I think once we do that, we still have, say, 2 weeks before we can give it to everybody.
80 00:10:42.230 ⇒ 00:10:45.890 Amber Lin: So… That’s… that’s the timeline there.
81 00:10:46.630 ⇒ 00:10:47.450 Pranav Narahari: Gotcha.
82 00:10:47.610 ⇒ 00:10:52.680 Pranav Narahari: And, this replicate and deploy embedding pipeline to Mashra.
83 00:10:53.120 ⇒ 00:10:56.910 Pranav Narahari: Does this have any dependence on the central docs?
84 00:11:00.010 ⇒ 00:11:02.460 Pranav Narahari: How are you guys thinking about this? Because I’m just…
85 00:11:02.770 ⇒ 00:11:07.529 Pranav Narahari: We have, like, embed and test with new central docs here.
86 00:11:08.010 ⇒ 00:11:22.469 Mustafa Raja: This was put in place before, Central Docs, improvement was discussed, even, so… The pipeline right now exists in N10, and we want to migrate it over to Master. So this is just saying that.
87 00:11:22.470 ⇒ 00:11:23.660 Samuel Roberts: Yeah, I think…
88 00:11:23.850 ⇒ 00:11:34.279 Samuel Roberts: there’s a world where this went, and then we did the central docs, and we still wanted to move the pipeline to Mastra out of NAD anyway, but those could be kind of combined in this…
89 00:11:34.630 ⇒ 00:11:36.090 Samuel Roberts: In this stage.
90 00:11:36.490 ⇒ 00:11:38.810 Pranav Narahari: Okay, I will… you know what, let me just do that.
91 00:11:38.810 ⇒ 00:11:42.330 Samuel Roberts: I mean, I wouldn’t necessarily say they’re exactly the same thing, but yeah, like…
92 00:11:42.330 ⇒ 00:11:44.600 Pranav Narahari: I think they should be done in tandem, though.
93 00:11:44.600 ⇒ 00:11:46.420 Samuel Roberts: That’s exactly how I’d say it, yeah.
94 00:11:47.470 ⇒ 00:11:48.240 Pranav Narahari: Okay.
95 00:11:49.460 ⇒ 00:11:53.909 Pranav Narahari: Let’s just do this all right now.
96 00:11:55.140 ⇒ 00:12:00.599 Amber Lin: Yeah, you wanna work… move the central dock sections up? I can do that.
97 00:12:01.280 ⇒ 00:12:03.710 Pranav Narahari: Oh yeah, that would be great, if you just move the whole thing up.
98 00:12:04.110 ⇒ 00:12:05.290 Amber Lin: Yeah, one sec.
99 00:12:07.300 ⇒ 00:12:07.890 Pranav Narahari: Thanks.
100 00:12:09.140 ⇒ 00:12:09.890 Amber Lin: Okay.
101 00:12:10.480 ⇒ 00:12:11.550 Pranav Narahari: Okay, perfect.
102 00:12:15.850 ⇒ 00:12:20.960 Pranav Narahari: Yeah, because I guess we didn’t see this being as tightly coupled.
103 00:12:22.910 ⇒ 00:12:23.560 Samuel Roberts: Goodbye.
104 00:12:23.560 ⇒ 00:12:32.659 Pranav Narahari: And yeah, just these last two parts are getting kind of similar to what we need to do for Central Duck, or at least it makes sense to do them with the other one in mind.
105 00:12:32.990 ⇒ 00:12:37.559 Pranav Narahari: Okay, migrate real to point to, BQ.
106 00:12:39.260 ⇒ 00:12:42.160 Amber Lin: I would say lower priority and later.
107 00:12:42.500 ⇒ 00:12:44.089 Pranav Narahari: 100%. Okay, cool.
108 00:12:44.230 ⇒ 00:12:49.299 Pranav Narahari: I’m actually just gonna take this off right now in terms of dates.
109 00:12:50.760 ⇒ 00:12:51.660 Samuel Roberts: Oh, yeah.
110 00:12:52.290 ⇒ 00:12:53.559 Pranav Narahari: And then,
111 00:12:54.600 ⇒ 00:13:02.910 Pranav Narahari: Casey, for this one, I think this is also a similar boat, right? Like, things are working right now, do you think this is gonna improve, latency?
112 00:13:03.180 ⇒ 00:13:04.899 Pranav Narahari: Or execution time, I should say.
113 00:13:06.570 ⇒ 00:13:13.759 Casie Aviles: Hmm… I think not a lot, since it’s just going to be moving it to…
114 00:13:13.930 ⇒ 00:13:16.740 Casie Aviles: the Google environment, right now it’s… Right.
115 00:13:17.630 ⇒ 00:13:22.990 Casie Aviles: Yeah, what I’m noticing is, like, the main reason, the main… or, like, the longest
116 00:13:23.300 ⇒ 00:13:26.560 Casie Aviles: The execution times is not really a… that…
117 00:13:26.690 ⇒ 00:13:29.099 Casie Aviles: In particular, it’s not super base.
118 00:13:29.480 ⇒ 00:13:32.619 Samuel Roberts: Yeah, it’s probably the generation of the SQL, right?
119 00:13:34.080 ⇒ 00:13:37.620 Pranav Narahari: Perfect. Yeah, let’s just remove that from the timeline right now.
120 00:13:37.810 ⇒ 00:13:45.600 Pranav Narahari: Okay, so this one is the most urgent one, I would say. Test run and QA for missing content.
121 00:13:45.770 ⇒ 00:13:53.609 Pranav Narahari: So, Mustafa, I think on this one, you did a little bit of QA, still needs my eyes to just do further QA on the mechanical central dock.
122 00:13:53.800 ⇒ 00:13:57.050 Pranav Narahari: What is the status for the other four?
123 00:13:58.650 ⇒ 00:14:09.680 Mustafa Raja: Yeah, for that, I, I asked for assistance on there, right? So I can take two of them, and then if someone else can take two of them, that would be nice.
124 00:14:11.570 ⇒ 00:14:13.000 Pranav Narahari: Perfect.
125 00:14:13.950 ⇒ 00:14:18.800 Pranav Narahari: Yeah, so the idea with this one is we want to get it over to them by end of day tomorrow. We kind of…
126 00:14:20.500 ⇒ 00:14:32.569 Pranav Narahari: like, I reiterated what kind of, I think we talked about, Amber, on Thursday, like… or actually, no, Monday, I think, in our call, we just said to them, we’ll send up… we’ll send over all five on Wednesday, right?
127 00:14:32.570 ⇒ 00:14:38.279 Amber Lin: Yeah, because they asked last week, and then they asked again this Wednesday, and Monday, so…
128 00:14:38.280 ⇒ 00:14:38.740 Pranav Narahari: Yeah.
129 00:14:38.740 ⇒ 00:14:43.700 Amber Lin: meeting with them Friday, so I… we need to give them something to prepare before we meet with them.
130 00:14:44.040 ⇒ 00:14:46.200 Pranav Narahari: I totally agree, yeah. So…
131 00:14:46.380 ⇒ 00:14:54.509 Pranav Narahari: Let’s get this done, yeah. Mustafa, if you get two done, I’ll work with, the team, or just I will get done… I’ll get the other two done.
132 00:14:55.360 ⇒ 00:14:56.350 Pranav Narahari: That works.
133 00:14:56.580 ⇒ 00:15:04.860 Amber Lin: Lawn and home improvement should be much smaller. Oh, I think there’s also commercial, so there will be… can I… can I read the departments you have listed?
134 00:15:05.420 ⇒ 00:15:11.969 Amber Lin: So there will be… Lawn, home improvement, and…
135 00:15:12.850 ⇒ 00:15:16.360 Amber Lin: Commercial, because we have mechanical and pest.
136 00:15:17.090 ⇒ 00:15:19.209 Pranav Narahari: Right, I think.
137 00:15:19.210 ⇒ 00:15:20.600 Amber Lin: 3…
138 00:15:20.600 ⇒ 00:15:23.759 Pranav Narahari: somewhere. Yeah, these are the ones right here.
139 00:15:24.200 ⇒ 00:15:27.940 Samuel Roberts: Yeah, we don’t need it for welcome teams, so just long…
140 00:15:28.020 ⇒ 00:15:31.929 Amber Lin: Commercial home improvement, so that will make it even faster. Just 3.
141 00:15:33.340 ⇒ 00:15:34.230 Pranav Narahari: Okay.
142 00:15:36.770 ⇒ 00:15:38.600 Amber Lin: Should I create sub-issues for them?
143 00:15:38.880 ⇒ 00:15:40.050 Amber Lin: Yeah, that would be nice.
144 00:15:48.210 ⇒ 00:15:52.780 Amber Lin: I… I can do that. Go ahead and continue on again, I can do that.
145 00:15:52.780 ⇒ 00:15:54.000 Pranav Narahari: Perfect, I appreciate that.
146 00:15:54.270 ⇒ 00:16:13.310 Pranav Narahari: So those are the three. Yeah, I can… I can take one of those. Mustafa, you can choose which one you… which two you want to take. And then what we’ll do, Mustafa, is we can just hop in a call tomorrow. Like, we’ll do our own QA, we’ll just, and then we can hop in a call tomorrow, do a final QA before I send out that message.
147 00:16:15.270 ⇒ 00:16:16.050 Mustafa Raja: Okay.
148 00:16:16.580 ⇒ 00:16:18.770 Pranav Narahari: Perfect, thank you.
149 00:16:19.920 ⇒ 00:16:33.240 Pranav Narahari: And embedding and testing will only happen after we get their feedback, so we’ll do our own internal QA. I’m gonna give them Thursday, Friday to do more QA. I’ll also ask them for a timeline on when they can give that back, because this embedding and testing
150 00:16:33.520 ⇒ 00:16:36.260 Pranav Narahari: with, new central docs, it’s gonna be depending on that.
151 00:16:38.170 ⇒ 00:16:41.799 Samuel Roberts: Yeah, there’ll still be work to do to, like, set up the pipeline and everything anyway.
152 00:16:43.210 ⇒ 00:16:46.130 Samuel Roberts: You know what I mean? Like, the… Oh, hold on.
153 00:16:46.280 ⇒ 00:16:49.709 Samuel Roberts: Yeah, I agree, I’m not disagreeing with what you’re saying.
154 00:16:50.340 ⇒ 00:16:50.730 Pranav Narahari: Exactly.
155 00:16:50.730 ⇒ 00:16:53.429 Samuel Roberts: Could happen early next week to set up the…
156 00:16:54.090 ⇒ 00:16:55.510 Samuel Roberts: You know, this kind of work.
157 00:16:55.810 ⇒ 00:16:57.209 Samuel Roberts: This one right here. Yeah.
158 00:16:57.450 ⇒ 00:16:59.320 Pranav Narahari: Yeah, perfect.
159 00:16:59.320 ⇒ 00:17:02.300 Samuel Roberts: But yeah, even if it’s, like, Wednesday or something, you know, this should be fine.
160 00:17:02.680 ⇒ 00:17:04.530 Pranav Narahari: Cool, yeah, I will…
161 00:17:04.859 ⇒ 00:17:11.720 Pranav Narahari: re… like, I’ll condense these into one. Or actually, it probably makes sense not for me to… let me leave them separate, but just, like…
162 00:17:12.160 ⇒ 00:17:17.190 Pranav Narahari: have this organized a little bit better. But okay, that’s good to know that…
163 00:17:17.589 ⇒ 00:17:20.109 Pranav Narahari: We can actually get started on this.
164 00:17:23.160 ⇒ 00:17:26.420 Pranav Narahari: We can get started on this a little bit earlier then, right?
165 00:17:27.339 ⇒ 00:17:28.060 Pranav Narahari: Because…
166 00:17:28.060 ⇒ 00:17:28.810 Samuel Roberts: Yeah, I don’t…
167 00:17:29.010 ⇒ 00:17:29.840 Pranav Narahari: Yeah…
168 00:17:30.080 ⇒ 00:17:35.440 Samuel Roberts: I don’t… think there’s anything… I don’t think this is necessarily a… Locker.
169 00:17:35.810 ⇒ 00:17:39.569 Samuel Roberts: I think it was kind of just scheduling, like.
170 00:17:39.720 ⇒ 00:17:47.130 Samuel Roberts: let’s get Andy set up, and then redo that, yeah. So I think that’s… Theoretically could be started.
171 00:17:48.770 ⇒ 00:17:49.830 Samuel Roberts: Whatever.
172 00:17:50.830 ⇒ 00:17:58.370 Pranav Narahari: Okay, yeah, so this is good to get started on. Okay, that’s good to know. Yeah, we might have too much on our plate right now to get started on.
173 00:17:58.370 ⇒ 00:18:01.540 Samuel Roberts: That’s… yeah, I’m not saying we need to, I’m just saying, like, it’s not being blocked, I think.
174 00:18:01.540 ⇒ 00:18:03.850 Pranav Narahari: Yeah, I think it shouldn’t have been linked to provide Andy.
175 00:18:03.850 ⇒ 00:18:04.740 Samuel Roberts: Yeah…
176 00:18:05.510 ⇒ 00:18:07.810 Pranav Narahari: Perfect. Okay,
177 00:18:08.370 ⇒ 00:18:15.949 Pranav Narahari: Roll out and test with clients, yep, makes sense. So this stuff right here, Central Dog Maintenance and Optimization,
178 00:18:16.860 ⇒ 00:18:19.169 Pranav Narahari: I know we’re doing a little bit of this, just…
179 00:18:20.320 ⇒ 00:18:27.470 Pranav Narahari: kind of, like, ad hoc, we don’t have, like, a script that does this automatically, for the restructuring. Like…
180 00:18:27.470 ⇒ 00:18:32.719 Amber Lin: Yeah, I feel like this is something we need to plan out, because we never…
181 00:18:32.980 ⇒ 00:18:45.289 Amber Lin: thoroughly thought this over. This is so that when they have new content, you said, maybe you have a form for them to fill in, so that we maintain, the structure and maintain the wording. So.
182 00:18:45.290 ⇒ 00:18:53.650 Amber Lin: Probably need a triage ticket, but maybe, like, 2 weeks later, we gotta plan it out, and then maybe redo these tickets, so…
183 00:18:53.650 ⇒ 00:18:54.100 Samuel Roberts: Yeah.
184 00:18:54.100 ⇒ 00:18:55.779 Amber Lin: Like, they might not be valid.
185 00:18:56.370 ⇒ 00:18:56.710 Pranav Narahari: Yeah.
186 00:18:56.710 ⇒ 00:19:01.299 Samuel Roberts: Yeah, I think these were kind of just placeholders for, like, a few ideas we had, but we hadn’t really…
187 00:19:01.560 ⇒ 00:19:03.509 Amber Lin: Yeah, scope them, really.
188 00:19:04.030 ⇒ 00:19:08.140 Amber Lin: I think I’ll just put, like, a planning ticket.
189 00:19:09.160 ⇒ 00:19:10.370 Amber Lin: on…
190 00:19:14.960 ⇒ 00:19:17.890 Pranav Narahari: Yeah, I think this makes a lot of sense.
191 00:19:19.190 ⇒ 00:19:21.300 Pranav Narahari: they are definitely interested in this. I think.
192 00:19:21.300 ⇒ 00:19:22.060 Samuel Roberts: Because…
193 00:19:22.060 ⇒ 00:19:38.530 Pranav Narahari: Great follow-up discussion to have with, for me to have with Janessa and Yvette, just about Janice and Yvette, combine their names, just based on how they see the central doc restructured, like, thing that we come up with.
194 00:19:40.370 ⇒ 00:19:48.939 Pranav Narahari: once we do that, I can get… I can be like, okay, moving forward, let’s work on how we make additions to this.
195 00:19:50.260 ⇒ 00:19:54.530 Pranav Narahari: Okay, I have, like, I have some ideas there, we can kind of go over that a little bit later.
196 00:19:57.870 ⇒ 00:20:02.729 Pranav Narahari: Okay, so, Casey, is this some… I think this is something you’re working through right now, right?
197 00:20:04.620 ⇒ 00:20:05.690 Pranav Narahari: Okay, perfect.
198 00:20:05.690 ⇒ 00:20:09.389 Casie Aviles: Why we were getting the errors on the channel.
199 00:20:09.960 ⇒ 00:20:11.890 Pranav Narahari: Okay, cool.
200 00:20:13.130 ⇒ 00:20:21.189 Pranav Narahari: what is, like, the status on this? What I want to do is what this, like, milestone I created here, just kind of showing a report.
201 00:20:21.300 ⇒ 00:20:25.890 Pranav Narahari: Do you think this week.
202 00:20:27.060 ⇒ 00:20:35.829 Casie Aviles: Yeah, I’m just running the… the script, really, and then I can share the CSV. That should be… and then I’ll… what I’ll do after that is I’ll just…
203 00:20:36.270 ⇒ 00:20:42.089 Casie Aviles: check which… which among those are in the triages, so… and then we can see, like, which…
204 00:20:42.340 ⇒ 00:20:45.909 Casie Aviles: We can see the comparison of the before and the after.
205 00:20:46.590 ⇒ 00:20:47.759 Pranav Narahari: Oh, perfect. And how long.
206 00:20:47.760 ⇒ 00:20:48.640 Casie Aviles: SCSV.
207 00:20:48.800 ⇒ 00:20:55.879 Pranav Narahari: a ton of, like, messages that you’re putting through the staging app. How long do you think that’s gonna take to run?
208 00:20:57.300 ⇒ 00:21:04.700 Casie Aviles: Hmm, I don’t have, like, a… it might take a couple of minutes, right now I’m still running the tests.
209 00:21:05.590 ⇒ 00:21:08.589 Casie Aviles: I wasn’t gonna say, like, hours or something like that.
210 00:21:09.590 ⇒ 00:21:17.759 Casie Aviles: Yeah, I’ll check again, but… I also have to worry about, like, rate limits, because if I… if I…
211 00:21:17.940 ⇒ 00:21:21.489 Casie Aviles: Run it, like, without any buffer, like, an interval.
212 00:21:21.910 ⇒ 00:21:28.190 Casie Aviles: It’s gonna… Returned, like, an error, and it says, like, it’s… we’re exceeding rate limits.
213 00:21:28.390 ⇒ 00:21:34.220 Casie Aviles: That’s the only issue there with having it run all through N8N.
214 00:21:35.070 ⇒ 00:21:41.399 Pranav Narahari: Gotcha. Yeah, so if that is the case, I would like to just run the triage tickets first.
215 00:21:41.560 ⇒ 00:21:44.999 Pranav Narahari: So if we can do, like, a pre-filter on what we send to…
216 00:21:45.580 ⇒ 00:21:48.440 Pranav Narahari: The staging app, that would be great.
217 00:21:48.920 ⇒ 00:22:00.329 Pranav Narahari: But if you feel like this is not gonna really be an issue, we can run through, like, all of them in a couple hours, we’re gonna have, like, a report, or we can start creating the report by end of day today, then, like, yeah, that’s great.
218 00:22:00.970 ⇒ 00:22:06.080 Casie Aviles: Yeah, it may take an hour if I’m going to add, like, intervals.
219 00:22:07.010 ⇒ 00:22:08.709 Pranav Narahari: Okay, yeah, an hour tonight.
220 00:22:09.720 ⇒ 00:22:13.110 Pranav Narahari: Perfect, thank you so much.
221 00:22:15.170 ⇒ 00:22:17.200 Pranav Narahari: Run triage issues with new work.
222 00:22:17.200 ⇒ 00:22:17.940 Casie Aviles: Those are…
223 00:22:18.420 ⇒ 00:22:19.440 Pranav Narahari: Duplicate, right?
224 00:22:20.670 ⇒ 00:22:21.550 Casie Aviles: Yeah.
225 00:22:22.190 ⇒ 00:22:29.149 Amber Lin: I was thinking, like, yes, it is the same function, but in the zip code one, we’re only looking at zip codes.
226 00:22:29.150 ⇒ 00:22:29.600 Samuel Roberts: Yeah.
227 00:22:29.600 ⇒ 00:22:46.229 Amber Lin: But the reason why we have this section of optimizing triage tickets is, can we make this process quicker for our team, and maybe for the clients as well in the future, because it takes up a lot of time, and our data quality depends on it.
228 00:22:48.650 ⇒ 00:22:49.200 Samuel Roberts: Yeah.
229 00:22:49.490 ⇒ 00:22:58.989 Amber Lin: So this is also, like, new planning we should do on how we want to improve it, but I agree, this is a… this ticket itself is a duplicate.
230 00:23:00.640 ⇒ 00:23:08.009 Pranav Narahari: Okay, I see what you’re saying. Maybe I’m not fully, actually…
231 00:23:08.160 ⇒ 00:23:13.009 Pranav Narahari: if… let me just kind of repeat what I’m thinking, like, this would be great for, which is…
232 00:23:13.090 ⇒ 00:23:29.389 Pranav Narahari: if we can run triage issues through the new workflow, then we can show that there’s no regression, and there’s only, like, progression of Andy, based on previous thumbs down, now becoming partially thumbs up, and all thumbs up from before are still thumbs up.
233 00:23:31.460 ⇒ 00:23:42.829 Amber Lin: I think that… yeah, I see that aspect as well. I think what I was trying to get at is the time it takes for us to address triage issues.
234 00:23:44.930 ⇒ 00:23:45.540 Pranav Narahari: Oh, I see.
235 00:23:45.540 ⇒ 00:23:51.800 Samuel Roberts: This has new workflow. We’re not necessarily meaning, the…
236 00:23:52.300 ⇒ 00:23:55.750 Amber Lin: I think it was just a workflow we developed to retest the old.
237 00:23:55.750 ⇒ 00:23:56.440 Samuel Roberts: Right.
238 00:23:56.440 ⇒ 00:23:59.429 Amber Lin: Right. That was it. That was as new as it gets.
239 00:23:59.790 ⇒ 00:24:02.560 Samuel Roberts: Yeah, I just wanted to make sure, but I think workflow here might be a little…
240 00:24:03.200 ⇒ 00:24:04.370 Pranav Narahari: Confusing.
241 00:24:05.240 ⇒ 00:24:08.140 Pranav Narahari: Gotcha. So there’s a new triage workflow.
242 00:24:10.330 ⇒ 00:24:11.850 Samuel Roberts: I think that’s what this…
243 00:24:12.840 ⇒ 00:24:20.149 Amber Lin: Let’s delete that and rename it. I think this is causing more confusion.
244 00:24:20.310 ⇒ 00:24:22.729 Amber Lin: than it should… Yeah, okay.
245 00:24:23.130 ⇒ 00:24:27.889 Amber Lin: So I… I think I’ll add, like, a spike to think about the…
246 00:24:28.640 ⇒ 00:24:33.250 Amber Lin: the time needed for triage the case, but it doesn’t have to be for now.
247 00:24:33.830 ⇒ 00:24:34.430 Pranav Narahari: Perfect.
248 00:24:35.420 ⇒ 00:24:39.870 Pranav Narahari: Okay, and then this one is another thing that they wanted to discuss.
249 00:24:39.980 ⇒ 00:24:45.369 Pranav Narahari: on Thursday, so I’ll… These will be for me.
250 00:24:45.690 ⇒ 00:24:50.819 Pranav Narahari: basically just, like, the weekend summary, like, we didn’t do these, like, the last two weeks, I think.
251 00:24:51.080 ⇒ 00:24:51.710 Samuel Roberts: Oh, yeah.
252 00:24:51.710 ⇒ 00:24:58.879 Pranav Narahari: And just, like, but the previous week, they really liked to just see, like, okay, what are the questions that were being asked this previous week? And then we can kind of…
253 00:24:59.160 ⇒ 00:25:01.049 Pranav Narahari: in that call a little bit, I think…
254 00:25:01.150 ⇒ 00:25:12.600 Pranav Narahari: the last couple have been, like, a little bit short. We can use, like, that remaining time to just kind of deep dive a little bit into the questions, what categories do we see? And then also, yeah, the thumbs-down analysis, like.
255 00:25:12.820 ⇒ 00:25:16.090 Pranav Narahari: What were the thumbs down? Let’s talk a little bit more about them.
256 00:25:16.090 ⇒ 00:25:24.189 Amber Lin: Cool. Do you think this is something we can get into real? Like, AI tagging classification of what the questions are?
257 00:25:24.510 ⇒ 00:25:28.259 Amber Lin: That will make it so much easier to present each week.
258 00:25:29.450 ⇒ 00:25:31.920 Pranav Narahari: Yeah, that’s true.
259 00:25:33.470 ⇒ 00:25:35.760 Amber Lin: So we’ll need to get it in Snowflake.
260 00:25:36.390 ⇒ 00:25:37.820 Amber Lin: And, like, a tap.
261 00:25:40.670 ⇒ 00:25:41.810 Pranav Narahari: Yeah. Yeah.
262 00:25:42.160 ⇒ 00:25:51.540 Pranav Narahari: I think endgame, that’s probably, like, the best thing to have. For this week, I’ll literally just, like, pull the info from Snowflake, and then just run the…
263 00:25:52.360 ⇒ 00:25:56.549 Pranav Narahari: the report manually, like, how I did it, like, 2 weeks ago.
264 00:25:56.690 ⇒ 00:25:58.829 Pranav Narahari: But, yeah, I agree, Amber, like…
265 00:25:58.980 ⇒ 00:26:01.500 Pranav Narahari: That’s just, like, one more thing that we could automate, so…
266 00:26:02.110 ⇒ 00:26:08.190 Samuel Roberts: I think that could easily be added to the Moscra flow, maybe, as, like, a… Parallel step that happens.
267 00:26:08.370 ⇒ 00:26:10.489 Amber Lin: To classify the question…
268 00:26:11.040 ⇒ 00:26:16.600 Samuel Roberts: And then that’ll just get tagged into the data that gets stored. That’s… will end up in real.
269 00:26:17.050 ⇒ 00:26:20.910 Samuel Roberts: So yeah, I don’t think that… I think, yeah, I think keep going the way you’re going, but that’s definitely something to…
270 00:26:21.470 ⇒ 00:26:22.210 Samuel Roberts: ads.
271 00:26:22.770 ⇒ 00:26:23.980 Pranav Narahari: Cool, I like that.
272 00:26:24.630 ⇒ 00:26:27.420 Pranav Narahari: Alright, ongoing maintenance, yup.
273 00:26:29.060 ⇒ 00:26:40.499 Pranav Narahari: Okay, so I feel pretty good about what we’re kind of gonna get done for the rest of this week. Mustafa, do you have time off as well? Casey, I know you’re gonna be out Thursday, Friday, right?
274 00:26:41.350 ⇒ 00:26:44.080 Casie Aviles: Wednesday, Thursday, Friday.
275 00:26:44.080 ⇒ 00:26:46.549 Pranav Narahari: Okay, you’re out tomorrow, too. Gotcha.
276 00:26:46.690 ⇒ 00:26:51.520 Pranav Narahari: Okay, so…
277 00:26:53.030 ⇒ 00:27:00.130 Pranav Narahari: The… yeah, the main two things is, like, the two reports. The… the zip codes report, and then also…
278 00:27:00.450 ⇒ 00:27:05.869 Pranav Narahari: well, yeah, the two reports and then also the central docs. That’s kind of the deliverables for this week.
279 00:27:06.060 ⇒ 00:27:19.399 Pranav Narahari: I think we’re in a good spot. I mean, we have a good amount of work to do for Central Docs, I think. But I think that’s taken care of now, we’ve figured out how we’re gonna split that up. And then Casey, I think by end of day today.
280 00:27:19.800 ⇒ 00:27:23.360 Pranav Narahari: that, the full, kind of.
281 00:27:23.850 ⇒ 00:27:32.589 Pranav Narahari: question or messages that are going through the staging master app should have been completed, and then me and you can work together about creating that report.
282 00:27:34.070 ⇒ 00:27:35.030 Casie Aviles: Okay.
283 00:27:35.550 ⇒ 00:27:36.210 Pranav Narahari: Cool.
284 00:27:36.690 ⇒ 00:27:46.740 Pranav Narahari: Alright, I think we’re in a pretty good spot, anything else, guys?
285 00:27:49.580 ⇒ 00:27:54.380 Samuel Roberts: I was gonna ask about transcripts, where that old workflow, work stream was.
286 00:27:55.090 ⇒ 00:28:10.380 Amber Lin: I think the last thing I’ve heard on the transcript side is that UTM wanted to create a scope of work for their transcript, so we’re gonna sign… I think we’re gonna sign something on it with the discovery deal, and then we will… Got it. Got it.
287 00:28:10.760 ⇒ 00:28:13.869 Samuel Roberts: Okay, cool, yeah, because I know I only loaded some, and I just wasn’t sure.
288 00:28:14.120 ⇒ 00:28:22.299 Amber Lin: Yeah, I think we’re… we’re good for now. One sec, let me read my notes. I had something…
289 00:28:22.650 ⇒ 00:28:25.760 Amber Lin: Oh, wanted to discuss…
290 00:28:25.960 ⇒ 00:28:36.910 Amber Lin: 1, 2, 3, 4. So, one, follow-up questions. I think we can see it under Andy Improvement Goals, if you scroll up a little bit.
291 00:28:37.160 ⇒ 00:28:47.750 Amber Lin: So, like, how… I think this is something that clients have been asking for a very long time, is can we have some probing questions?
292 00:28:47.940 ⇒ 00:28:48.700 Pranav Narahari: Yeah.
293 00:28:48.700 ⇒ 00:28:49.720 Amber Lin: Something is…
294 00:28:50.200 ⇒ 00:28:57.529 Amber Lin: it doesn’t know what to give. They give bad questions, but unfortunately, they don’t know what to ask.
295 00:28:58.010 ⇒ 00:28:58.510 Samuel Roberts: Right.
296 00:28:58.510 ⇒ 00:28:59.220 Amber Lin: Or there’s…
297 00:28:59.220 ⇒ 00:29:04.670 Samuel Roberts: a few times where it asks follow-up, like, do you mean inspector, or.
298 00:29:04.670 ⇒ 00:29:08.249 Amber Lin: Yeah, so we have it on some of the zip codes, which is really helpful.
299 00:29:08.250 ⇒ 00:29:08.720 Samuel Roberts: Yeah.
300 00:29:08.720 ⇒ 00:29:10.459 Amber Lin: have an overarching.
301 00:29:10.640 ⇒ 00:29:13.939 Samuel Roberts: Which I feel like could be an easier win for us. I think so.
302 00:29:13.940 ⇒ 00:29:15.230 Amber Lin: improve accuracy.
303 00:29:18.350 ⇒ 00:29:19.139 Casie Aviles: So, it’s…
304 00:29:19.360 ⇒ 00:29:25.380 Pranav Narahari: implementation for that with zip codes, how easy is it to just kind of make it more broad scope?
305 00:29:27.420 ⇒ 00:29:31.370 Pranav Narahari: Because right now, we have that for zip codes, right? But… We don’t have a.
306 00:29:31.370 ⇒ 00:29:32.090 Casie Aviles: Yeah.
307 00:29:34.950 ⇒ 00:29:44.209 Casie Aviles: Yeah, because the way we’ve been dealing with follow-up questions is whenever, like, it’s kind of like a special case each time where
308 00:29:44.820 ⇒ 00:29:47.680 Casie Aviles: Or, like, a specific case where we…
309 00:29:48.100 ⇒ 00:29:54.460 Casie Aviles: prompt Andy, basically, to… if the user asks about this, then ask a follow-up question, so that’s…
310 00:29:54.860 ⇒ 00:30:00.930 Casie Aviles: Kind of, you know, very specific, but it’s not like… Yeah.
311 00:30:00.930 ⇒ 00:30:04.039 Samuel Roberts: Are you talking about how it’s done currently in N8N?
312 00:30:04.430 ⇒ 00:30:09.100 Casie Aviles: Yes, yes. Yeah. So it’s just prompting… Mainly prompting.
313 00:30:09.100 ⇒ 00:30:18.870 Amber Lin: It’s kind of like a decision tree of if asked about, say, attributes, asked about these possible branches of things, and then this is how we progress.
314 00:30:20.700 ⇒ 00:30:24.669 Pranav Narahari: Gotcha. Yeah, that makes sense. It seems like,
315 00:30:25.310 ⇒ 00:30:32.760 Pranav Narahari: So, okay, so you mentioned how that is how we’re doing it for N8M. For Mashra, it’ll probably be something similar, right? Just not, like…
316 00:30:33.350 ⇒ 00:30:48.339 Pranav Narahari: a module. Like, we’re just asking it a question to be like, if in this topic we ask a follow-up question, seems like, yeah, just updating that system prompt to just include also the additional topics that we want to include,
317 00:30:48.960 ⇒ 00:30:49.890 Pranav Narahari: And…
318 00:30:50.080 ⇒ 00:30:59.799 Pranav Narahari: So it takes two things, I guess. The implementation seems pretty easy, the actual research of, like, okay, which things do we want to ask follow-up questions to? That’s the main component of this.
319 00:31:00.210 ⇒ 00:31:01.529 Pranav Narahari: She already answered, though?
320 00:31:03.480 ⇒ 00:31:21.479 Samuel Roberts: Yeah, I mean, it might even be something as simple, possibly, as just telling the system, or in the system prompt saying, you know, like, be 100% certain about what you answer. If you’re not 100% certain, ask… ask a clarifying question, and it might… with the newer models, especially now that we’re moving on from 4.0, it might be able to…
321 00:31:21.680 ⇒ 00:31:31.609 Samuel Roberts: do that on its own without us having to specify, like, here’s the follow-up questions we want you to ask. That being said, if there are specific ones, we definitely should get those in there.
322 00:31:31.610 ⇒ 00:31:43.909 Amber Lin: Yeah, agreed, because for us to add in every single one of, hey, for this question, here are the follow-ups, I think it’s hard for us to maintain. Yeah, definitely. I was wondering if there’s something of…
323 00:31:44.010 ⇒ 00:31:51.779 Amber Lin: It can gather from the context we have, because we have a lot of context of what type of thing deserves follow-up questions.
324 00:31:51.960 ⇒ 00:31:54.790 Amber Lin: But then also, Thank you.
325 00:31:55.220 ⇒ 00:31:55.810 Amber Lin: A.
326 00:31:56.720 ⇒ 00:31:57.850 Amber Lin: Sorry.
327 00:31:58.200 ⇒ 00:32:12.050 Amber Lin: I worry that if we ask follow-up questions on every single thing, then it’s not going to be as real-time or as fast. That’s the downside of follow-up questions, but I think it will improve quality significantly.
328 00:32:14.820 ⇒ 00:32:25.980 Amber Lin: I guess one of the things is… I know in the new Central Docs we’re creating, there’s, like, a decision tree section of, like, decision logic.
329 00:32:26.120 ⇒ 00:32:37.770 Amber Lin: Is that what this is for? Janice asked me about it, too. I was like, is this for how we will route probing questions? Like, what is the decision logic section for?
330 00:32:42.390 ⇒ 00:32:43.360 Samuel Roberts: Take that.
331 00:32:43.560 ⇒ 00:32:44.400 Samuel Roberts: Yeah.
332 00:32:44.580 ⇒ 00:32:46.130 Samuel Roberts: I think that was to help…
333 00:32:46.340 ⇒ 00:32:48.270 Samuel Roberts: What was that to help? Andy?
334 00:32:48.770 ⇒ 00:32:49.699 Mustafa Raja: Yeah… I think it was.
335 00:32:49.700 ⇒ 00:32:55.600 Samuel Roberts: Yeah, I’m sorry, I don’t have it in front of me right now, but, go ahead, Mustafa.
336 00:32:55.600 ⇒ 00:33:00.360 Mustafa Raja: It was, it was just to, support, existing stuff.
337 00:33:00.690 ⇒ 00:33:08.030 Mustafa Raja: That, what’s it called? Agent couldn’t, think a better, place for.
338 00:33:08.900 ⇒ 00:33:16.009 Mustafa Raja: And then also how we are making some of the decisions. This was based on PEST document.
339 00:33:16.810 ⇒ 00:33:23.460 Mustafa Raja: It had best document in context, and that’s… that’s how it suggested all of those sections.
340 00:33:24.330 ⇒ 00:33:27.380 Amber Lin: I see. Do you guys think having…
341 00:33:27.380 ⇒ 00:33:27.880 Mustafa Raja: Hello!
342 00:33:27.880 ⇒ 00:33:29.349 Amber Lin: Oh, sorry, go ahead.
343 00:33:29.680 ⇒ 00:33:37.059 Mustafa Raja: Yeah, I did take a look back at it, and I’m wondering if we really need this… need that section at all.
344 00:33:37.060 ⇒ 00:33:40.440 Amber Lin: I see. I don’t see a Central Doc specific.
345 00:33:40.790 ⇒ 00:33:44.219 Mustafa Raja: content in there. Oh.
346 00:33:44.220 ⇒ 00:33:51.119 Amber Lin: Yeah. Janiece was very confused, so I wanted to bring it up before you guys, like, generate new stuff.
347 00:33:52.140 ⇒ 00:33:59.090 Mustafa Raja: So, I think we can… we can still have everything after… even after removing that, you know.
348 00:34:01.090 ⇒ 00:34:02.670 Mustafa Raja: Cool.
349 00:34:03.020 ⇒ 00:34:11.009 Mustafa Raja: So, Pranav, is that chat, from, from the Central Dog development? This one, yeah.
350 00:34:11.010 ⇒ 00:34:11.819 Pranav Narahari: Yeah, this is the one
351 00:34:12.880 ⇒ 00:34:17.880 Pranav Narahari: Your channel. This is the one I linked… okay, okay, okay. I think this is… this is a somewhat older one.
352 00:34:18.190 ⇒ 00:34:22.910 Mustafa Raja: But I’ll see… I’ll see if I can, share a newer one or not.
353 00:34:23.270 ⇒ 00:34:25.409 Mustafa Raja: But this should have context on it.
354 00:34:27.120 ⇒ 00:34:29.430 Pranav Narahari: Okay, yeah, there’s decision logic here.
355 00:34:31.620 ⇒ 00:34:32.520 Pranav Narahari: Okay.
356 00:34:38.110 ⇒ 00:34:46.779 Pranav Narahari: Okay, so… For probing questions, is that something that we want to create a ticket for right now? Like…
357 00:34:47.929 ⇒ 00:34:49.130 Pranav Narahari: I’m trying to think, like.
358 00:34:49.130 ⇒ 00:34:50.920 Mustafa Raja: Do you think probing questions are good?
359 00:34:52.300 ⇒ 00:34:53.470 Pranav Narahari: My, my…
360 00:34:53.760 ⇒ 00:35:06.299 Pranav Narahari: my concern with probing questions is that we’ll just be asking generic probing questions that aren’t going to be helpful, or we ask just, like, questions that are just like, why are you asking me this? And…
361 00:35:06.520 ⇒ 00:35:17.419 Pranav Narahari: another thing is… is that, yeah, if we think about who is going to be using Andy right now, is… it’s going to be people that are on a call, like, we want to get them the quickest.
362 00:35:17.900 ⇒ 00:35:24.889 Pranav Narahari: We want to get them an answer as quick as possible, so we don’t want to be asking them unnecessary probing questions.
363 00:35:25.220 ⇒ 00:35:40.350 Pranav Narahari: So what I kind of want to first see is just, like, where is the accuracy at without probing questions? And we’ve done a lot of updates currently, or we’re in the process of doing a lot of updates currently for accuracy. Let’s see what metrics, what,
364 00:35:40.660 ⇒ 00:35:49.779 Pranav Narahari: quantitative values we have there for accuracy. And then we make the assessment, does it make sense to increase the number of probing questions we ask?
365 00:35:49.950 ⇒ 00:35:52.530 Pranav Narahari: I also talked to Janiece…
366 00:35:52.750 ⇒ 00:36:08.279 Pranav Narahari: sorry, I talked to Janiece last week, but I talked to Yvette on Monday, to discuss probing questions specifically, and yeah, I brought up… she was talking about how, like, probing questions are great, and that she wanted them, but at the same time, she was also talking about
367 00:36:08.470 ⇒ 00:36:15.179 Pranav Narahari: we need that… we need Andy to be faster at some points, and I was like, okay, that’s two different things.
368 00:36:15.360 ⇒ 00:36:16.440 Pranav Narahari: And I think…
369 00:36:16.800 ⇒ 00:36:26.949 Pranav Narahari: where we ended up with was just like, okay, probing questions is not a priority as of right now. I think the main priority is just the things that we’re working on.
370 00:36:27.180 ⇒ 00:36:34.020 Pranav Narahari: which are going to be helping with accuracy, reducing execution time. Then we’ll have probing questions as, like.
371 00:36:34.330 ⇒ 00:36:38.770 Pranav Narahari: Something that we can implement going forward to further increase accuracy.
372 00:36:38.770 ⇒ 00:36:41.170 Amber Lin: Good point. I like how you put that.
373 00:36:42.320 ⇒ 00:36:43.020 Pranav Narahari: Cool.
374 00:36:45.300 ⇒ 00:36:46.980 Pranav Narahari: Alright,
375 00:36:50.790 ⇒ 00:36:56.250 Pranav Narahari: Let’s see, I feel… I feel pretty good, guys. Anything… anything else?
376 00:36:58.220 ⇒ 00:37:01.949 Pranav Narahari: Amber, I think you said you had a couple other things you wanted to talk about, or was that everything?
377 00:37:01.950 ⇒ 00:37:17.069 Amber Lin: The other thing was, like, aliases, and how we respond to specific aliases, and how we keep that up to date. Like, for example, for them can mean after hours, and we don’t…
378 00:37:17.850 ⇒ 00:37:22.389 Amber Lin: I don’t think we actively… we had a log, but we haven’t been updating it.
379 00:37:22.530 ⇒ 00:37:41.129 Amber Lin: And then for QC, for example, they want a specific response of, hey, send service managers if it’s about QC. So, if we’re little, small logics like that, how do we handle it, and how… are we just going to keep doing manual maintenance on those things, or…
380 00:37:41.700 ⇒ 00:37:46.480 Amber Lin: How is it… how’s the workflow gonna be like in the future?
381 00:37:46.660 ⇒ 00:37:48.519 Amber Lin: With a new app and all that.
382 00:37:52.320 ⇒ 00:37:53.540 Casie Aviles: So, it def…
383 00:37:53.540 ⇒ 00:38:02.229 Pranav Narahari: The number of aliases, I guess. I’m not… what is the number of aliases? Like, is this, like, are we getting a new one, like, every day? Like…
384 00:38:02.620 ⇒ 00:38:03.950 Pranav Narahari: every week.
385 00:38:04.070 ⇒ 00:38:10.420 Pranav Narahari: If it’s just one every week, it seems like something you throw on a system prompt to say, like, this means this.
386 00:38:11.740 ⇒ 00:38:15.150 Pranav Narahari: Am I wrong there? Do you think that’s gonna actually not perform?
387 00:38:15.150 ⇒ 00:38:20.480 Amber Lin: I think we’re at, like, one every day, probably… probably it’ll be more when we start out, but…
388 00:38:20.710 ⇒ 00:38:22.210 Samuel Roberts: Yeah, I think it’ll level off.
389 00:38:22.210 ⇒ 00:38:22.950 Amber Lin: week.
390 00:38:22.950 ⇒ 00:38:27.120 Samuel Roberts: Yeah, I think that’s stuff that can go in the system prompt for now, and if we need to add
391 00:38:27.720 ⇒ 00:38:29.230 Samuel Roberts: Something more robust.
392 00:38:29.490 ⇒ 00:38:37.750 Samuel Roberts: if it’s a lot more over time, but I think it will… like, there’s probably a bunch that aren’t recognized, and once we start recognizing those, it’ll be less of a…
393 00:38:38.210 ⇒ 00:38:40.469 Samuel Roberts: Less frequent, that there’s new ones.
394 00:38:40.470 ⇒ 00:38:40.820 Amber Lin: Okay.
395 00:38:40.940 ⇒ 00:38:42.229 Pranav Narahari: And just so I think…
396 00:38:42.230 ⇒ 00:38:42.810 Samuel Roberts: Yeah.
397 00:38:42.810 ⇒ 00:38:44.169 Pranav Narahari: So I know for sure, like.
398 00:38:44.380 ⇒ 00:38:53.130 Pranav Narahari: what this alias concern is, is that in the central doc, we call something different than what they’re saying in ANDI, right? And they mean the same thing.
399 00:38:55.680 ⇒ 00:39:01.699 Amber Lin: Yeah, I think for the most part. Especially if they’re typing really fast, they don’t type the whole word sometimes.
400 00:39:01.700 ⇒ 00:39:07.650 Pranav Narahari: Okay, and so we just want to be able to accommodate those abbreviations. Okay.
401 00:39:07.830 ⇒ 00:39:16.610 Samuel Roberts: Yeah, I think there’s definitely something to think about, like, what’s a generic one, like, after hours, maybe, and what might be more specific that should be in the central doc, maybe?
402 00:39:17.740 ⇒ 00:39:22.760 Samuel Roberts: But I think, for the most part, system prompt is the way to go, if they’re broad.
403 00:39:24.780 ⇒ 00:39:34.560 Pranav Narahari: Yeah, sounds good. And then that, too, seems like not super high priority right now. It’s another thing that is going to bump up accuracy, for sure.
404 00:39:35.480 ⇒ 00:39:41.859 Pranav Narahari: So… We can probably get started on that next week, for this week, I think.
405 00:39:42.030 ⇒ 00:39:44.180 Pranav Narahari: I think we have our hands full, and I think…
406 00:39:44.470 ⇒ 00:39:48.160 Pranav Narahari: We’ll be able to deliver on a lot of stuff, which is… which is great.
407 00:39:51.880 ⇒ 00:39:52.820 Pranav Narahari: Cool, guys.
408 00:39:55.920 ⇒ 00:39:58.070 Samuel Roberts: Yeah, I feel pretty good.
409 00:39:59.690 ⇒ 00:40:00.800 Pranav Narahari: Think we’re all set.
410 00:40:01.410 ⇒ 00:40:01.960 Samuel Roberts: Okay.
411 00:40:02.700 ⇒ 00:40:03.170 Amber Lin: Cool.
412 00:40:03.520 ⇒ 00:40:04.800 Amber Lin: Yeah.
413 00:40:04.840 ⇒ 00:40:10.759 Pranav Narahari: One thing is with Linear, haven’t been able to tackle that yet. I think for me…
414 00:40:11.310 ⇒ 00:40:15.179 Amber Lin: I can help with that, I just also haven’t put my hands on it.
415 00:40:15.760 ⇒ 00:40:25.610 Pranav Narahari: Okay, yeah, all good. I think, Amber, we’re kind of synced based on this meeting, like, what we need to get done this week, the tickets, that should be kind of, like, reflecting that, but…
416 00:40:26.110 ⇒ 00:40:30.620 Pranav Narahari: Yeah, probably going forward, just like, we don’t want to just be scoping week-to-week like this, probably…
417 00:40:31.250 ⇒ 00:40:36.560 Pranav Narahari: starting next week, or later this week, Amber, we can talk about How we can just, like.
418 00:40:37.230 ⇒ 00:40:42.299 Pranav Narahari: restructure linear, make sure everything is, looking tidy there.
419 00:40:42.300 ⇒ 00:40:42.920 Amber Lin: Yeah.
420 00:40:43.090 ⇒ 00:40:46.709 Amber Lin: Cool, sounds good, because our timeline kept moving.
421 00:40:46.710 ⇒ 00:40:49.059 Samuel Roberts: Yeah. In the past, so hopefully we can…
422 00:40:49.160 ⇒ 00:40:50.729 Amber Lin: Have that more settled.
423 00:40:51.040 ⇒ 00:40:51.840 Samuel Roberts: Yeah.
424 00:40:52.170 ⇒ 00:40:52.980 Pranav Narahari: Alright.
425 00:40:53.240 ⇒ 00:40:53.970 Pranav Narahari: Awesome.
426 00:40:54.960 ⇒ 00:40:56.220 Samuel Roberts: Right. Michelle?
427 00:40:56.870 ⇒ 00:40:58.039 Pranav Narahari: Yeah, thanks guys.
428 00:40:59.970 ⇒ 00:41:00.590 Samuel Roberts: See you later.
429 00:41:00.910 ⇒ 00:41:01.550 Pranav Narahari: Sip.