Call —
Transcript batch
0:00 Caller: This call will be recorded so we can focus on you, not taking notes.
0:00 You: This call will be recorded.
0:05 You: Hello.
0:08 You: How about now?
0:13 You: Hello?
0:15 Caller: Hi, Jed, it's Ashley Clancy with What Not. How are you?
0:19 You: I'm well in yourself.
0:21 Caller: I'm doing very well. Thank you for asking. Is now still a good time for you to chat?
0:27 Caller: All right. Excellent. And thank you again for taking the time to speak with me. I'd love to use this time just to learn a little bit more around your background. And then we can discuss the team a bit more. How does that sound? All right. Excellent. So I'll let you take it away. I'd love to learn a bit more around your career trajectory and what you've been working on recently.
0:30 You: Sure.
0:44 You: Sure.
0:45 You: So the overarching theme of my career has been automation.
0:57 Caller: Thank you.
1:00 You: building tools that liberate people from monotonous tasks, starting in Wall Street on the equity derivatives desk at Sochgen, moving to marketing technology at a place called ad quant media and content IQ.
1:14 You: Moving on from there to big tech at Spotify, measuring the effectiveness of different, what are called above-the-line marketing campaigns, things like TV commercials, billboards, stuff you can't really measure the plate, the way you can.
1:27 Caller: Thank you.
1:30 You: at a place like meta, where at meta, I was responsible for the telemetry for Instagram's notification feeds.
1:37 You: So all events, all clicks, taps, impressions, came, passed through my telemetry and ultimately went from data into information.
1:52 You: And most recently at Roe, I was brought on to help the underwriting team.
1:56 You: And after increasing their underwriting capacity 400X, leadership asks, what else can you do?
1:57 Caller: Yeah, you know, I'd love to learn, and I'd love to learn, um, and I'd love to learn, you know,
2:03 You: And now every leader at Roe consumes something from my automata regularly.
2:13 You: Is there any place you'd like to elaborate?
2:15 You: Like me to elaborate?
2:21 Caller: maybe around an example project. And I'd love to learn what the business problem or
2:26 You: So one feature that came out fairly recently at Roe was receipt inbox.
2:27 Caller: challenge was, technically, you know, speaking specifically what you built, and then the end
2:33 Caller: results or impact that was measured for the business.
2:42 You: The idea is when expenses are incurred, you have to get the receipt and submit it into the platform, so that can be synchronized back to the accounting integration.
2:52 You: And at the time, users were required to...
2:56 You: to manually map the uploaded receipts to the transactions responsible for them.
2:57 Caller: Thank you.
3:01 You: So I was tasked with figuring out how to reduce that burden.
3:07 You: Ultimately, I ended up using a combination of heuristics, some machine learning, and LLMs,
3:13 You: to take uploaded receipts, identify specific metadata about them, date, the last four of the card,
3:21 You: the amount, things like that, and then use that to narrow down a list of probable transactions
3:26 You: along with the user responsible for those transactions.
3:27 Caller: Thank you.
3:30 You: And as a result, I think it was like 96% of receipts were now processed automatically.
3:37 You: And the remaining 4%, a large part of them were not even receipts to begin with.
3:42 You: So in practice, the actual accuracy is likely higher.
3:47 You: The ultimate business impact was this new feature that people who are incurring charges in the field now appreciated,
3:55 You: mainly because their expenses were tracked more automatically for them,
3:57 Caller: Excellent.
4:00 You: and they were able to get reimbursed faster.
4:03 You: And this applied to the majority of rows customers who have some sort of row card.
4:09 You: The underlying infrastructure was built cheap enough that the cost of any incremental submitted receipt
4:17 You: was on the order of pennies, far cheaper than a human analyst doing that in the background.
4:25 You: So three, four weeks ago now, a founder found my GitHub
4:27 Caller: Excellent, excellent. Thank you for sharing. All right. And now switching gears for a moment. You know, as far as why you're looking for a moment, you know, as far as why you're looking for a switch, I guess, what's motivating you for a change at this time?
4:45 You: and cold emailed me suggesting I interview for a role with their firm.
4:51 You: That's like, that process is still live, but I figure, but I figure if they're looking,
4:55 You: maybe it's time to reassess how other companies value my accumulated skills, knowledge, and experience.
4:57 Caller: Okay, cool. All right. All right. All right. And then. All right. And then. And then.
5:01 You: What made the original whatnot role not stand out is it was a subset of things that I've only experimented with,
5:09 You: not really what I pushed into production.
5:11 You: The one that my agents found was much closer to the kinds of things that are currently cutting edge for me,
5:17 You: and at least much closer to my present day to day.
5:25 You: So three things that I look for in any role.
5:27 Caller: You know, as far as, you know, that next role for you, I guess, I'd love to dive a little deeper here and to learn a bit more around what excites you next, what you'd like to be working on in your day-a-day.
5:43 You: The first is data is both present and necessary.
5:47 You: Instinct and insight are good, but I'm of the opinion they must always be backed by data.
5:52 You: The second is the second is the relative.
5:55 You: to pursue what I genuinely think is the best solution to a problem.
5:57 Caller: Okay.
5:59 You: My career is quite diverse, and it's getting longer, so I want to be able to bring forward that earned skills, knowledge, and experience into the next role.
6:04 Caller: Okay. And then as far as what not you know what the product. Like, I guess what it's
6:07 You: And the last is a place for the phrase, that's not my job, doesn't exist.
6:11 You: I see that as a sign of bureaucracy and I try to avoid it.
6:25 Caller: I guess what it kind of excites you about the company, the product, the mission, things of that nature.
6:25 You: So I liken what not to an evolution of what TikTok shop meant to be.
6:39 You: And this experiential online shopping, at least from the e-commerce folks whom I'm connected with,
6:46 You: suggested this is no longer the future, it's the present, meaning I'm in the past.
6:51 You: So being able to arrive closer to the forefront of this.
6:55 Caller: Sure. I love that analogy. It's whatnot, like and whatnot to the evolution that TikTok shop, like, wanted to, wanted to see. That's funny. All righty. And then just a couple of housekeeping questions. So location, we're based in the greater New York area.
6:55 You: this space, while it's still relatively unknown, at least among my peers, was made it worth exploring.
7:25 Caller: We're targeting or difference. Okay. Yeah, excellent. That, you know, that works for whatnot. We're welcome. We're open, rather, to folks working remotely. We just require employees to live 50 miles, within 50 miles of one of our office hubs. Our New York City hub is in Hudson Square. And if you're in New York City property, you're well within 50 miles of the hub. So we're all set there.
7:25 You: Yes, I'm in New York City proper.
7:52 Caller: And then as far as far as,
7:55 Caller: as interviews. You mentioned you have an ongoing process, I guess, like, what stage are you at, like, how far along are you with other, other processes?
7:55 You: I'm all over the place.
8:06 You: Some are the recruiter stage similar to here now.
8:06 Caller: Okay.
8:09 You: Others are second and third round technical or other kinds of interviews.
8:11 Caller: Okay. Okay. All right. And then your timeline, I guess, when, you know, assuming you've received an offer, when would you be looking to join
8:25 Caller: and start your next opportunity.
8:25 You: So that one's on the order of possibly three to four weeks after accepting an offer,
8:33 You: only because every leader consumes something from me.
8:36 You: So it'll take a while to transfer it all to their respective teams.
8:41 Caller: Okay. All right. And then compensation, are there any comp expectations that you'd like to discuss today?
8:53 You: Nothing specific.
8:55 Caller: Sure. All right. Excellent. Well, all right. Excellent. Well, thank you for chatting through some of those follow-up questions with me. So now I'll, you know, discuss a little bit more just around the team and ML here at whatnot. So essentially,
8:55 You: we move closer to an offer, I'd like to get more diligence into the cap table and the preference
9:01 You: stack if there is one. What the budget listed in the job and the role was within range.
9:23 Caller: machine learning here breaks down into three groups you can imagine. So we can imagine that we have two
9:25 You: Thank you.
9:31 Caller: groups that are more focused on the application side of things. This is our fraud, you know, fraud. They're
9:38 Caller: handling, you know, exactly what that sounds like, financial ML, refund abuse, you know, things of that
9:42 Caller: nature. And then we have our discovery group. This is really anything that is touching the application
9:48 Caller: directly. So this is, these are things like, you know, ads, growth, seller tools.
9:53 Caller: trust and safety, buyer tools, you know, things of that nature. And then we have the
9:55 You: Thank you.
9:57 You: Thank you.
9:59 Caller: ML core team. And so this ML core team is owning everything that the two app teams, you know,
10:07 Caller: application groups don't handle as well as all of the infrastructure model training and serving
10:13 Caller: for all teams at whatnot. They own the AI platform and the production-facing LLM infrastructure
10:21 Caller: as well. You can almost imagine them as a horizontal team kind of working across the whole
10:27 You: Thank you.
10:28 Caller: landscape of whatnot, whereas the application, you know, the teams within the, you know,
10:29 You: Thank you.
10:34 Caller: the more application focused are pretty siloed into that one specific business area,
10:39 Caller: business, you know, set of business challenges. And so this ML Corps team has a few
10:45 Caller: different initiatives, right? So they're working on standardizing real-time features.
10:51 Caller: serving for all models. We're moving away from tree-based models on CPUs to large
10:58 Caller: deep learning models on GPUs. And we're migrating off of StageMaker over to Kubernetes for model
10:59 You: So I'm
11:05 Caller: serving. And we're working on building, you know, we have a big undertaking or trying to stand up
11:12 Caller: an in-house feature store from the ground up as well. So they've got a lot of projects going on
11:19 Caller: with an ML Corps. I'll stop there. I'd love to hear just your thoughts and how that's
11:26 Caller: resonating.
11:28 Caller: I believe so, yes. I believe so, yes. Like I, that's, that's, that's, let me, let me, unless I was mistaken.
11:29 You: you reached out for back in May.
11:32 You: I mentioned the I responded more recently about the LLM platform engineer role.
11:37 You: Is that still in the same family?
11:39 You: Because what's still in the same family.
11:48 You: Because what you described sounded like MLOPS, not quite LLM wrangling.
11:48 Caller: was mistaken.
11:49 Caller: Give me, yeah, that's actually a fair point.
11:52 You: Rangling.
11:59 You: Of course.
12:00 Caller: Give me one, one moment here, Jed.
12:03 Caller: That's my system is, of course, taking forever to think like it always does.
12:06 Caller: Okay.
12:07 Caller: Yeah, so this, yeah, LLM platform engineer, this is still, this is, you know, within our, you know,
12:16 Caller: within our ML core teams, like I mentioned, our ML core teams own all of the infrastructure
12:24 Caller: that power the MLMs here. That's what this LL platform engineer role is.
12:28 Caller: It is. It is that, I would say, I would say it's that also. There are machine learning engineers
12:29 You: Understood. So it's not necessarily tuning GPUs or setting up post-training learnings.
12:36 You: It's a little more nuanced than that.
12:40 You: Understood. Okay.
12:43 Caller: are end-to-end here.
12:46 You: It still aligns.
12:46 Caller: Yeah. Yeah. How, you know, how does that kind of resonate feel?
12:50 You: It's still aligns.
12:53 Caller: Okay. Okay. Excellent. Excellent. So what, you know, so next from here, next steps, what I will do is share your
12:55 You: It's how I put this.
12:57 You: In the sphere of skills.
12:58 You: It's...
12:59 You: It's focused on a different set than I had originally prepared for, but still within my capability.
13:16 Caller: profile with the leaders from our ML Corps group. They're usually pretty responsive with
13:23 Caller: feedback. I would expect to have an update from them either by the end of the day today or, you know,
13:28 Caller: likely first thing, first thing tomorrow. If the, you know, if they're like, yes, let's proceed,
13:29 You: Thank you.
13:31 You: Thank you.
13:34 You: Thank you.
13:35 Caller: as far as the first step in the interview process, that would be a 30 to 45 minute interview
13:36 You: Thank you.
13:38 You: Okay.
13:43 Caller: with the leader from our ML Corps team, where you would discuss a bit more around, you know,
13:45 You: Good to know.
13:50 Caller: the team, you know, their scope and projects, and then you would deep dive into some of your
13:56 Caller: recent, relevant work and impact as well.
14:03 Caller: Yeah, just so I guess, just so you know what, what kind of to expect. So essentially from here,
14:08 You: Sounds good.
14:08 Caller: I will share your profile with our leaders. Once we have an update, we will, we will, we will,
14:13 Caller: we will reach out to you accordingly.
14:15 Caller: If the team is looking to kick off the interview process, it would start with that initial hiring manager interview.
14:23 Caller: All right.
14:24 You: Good.
14:25 You: Good.
14:26 You: What is the
14:26 Caller: Any questions I could answer for you?
14:30 You: What is the anticipated time commitment for this?
14:30 Caller: Sure.
14:36 You: I asked
14:37 You: because.
14:38 You: have multiple processes in flight, so I don't know to what extent.
14:42 You: I can ask what not to compress or them to extend.
14:43 Caller: Yeah.
14:44 Caller: Yeah, no, 100%.
14:47 Caller: So after the hiring manager interview would be our system design that is 60 minutes long, from there, we move
14:55 Caller: move to the coding interview.
14:57 Caller: You have a choice of either doing this live with.
14:59 Caller: live with, you know, live with whatnot, it would be 60 minutes if it's live, or you can do the
15:06 Caller: take-home coding. You know, you can complete it. The only ask is that you return it back to us
15:08 You: Thank you.
15:09 You: Thank you.
15:11 You: Thank you.
15:11 Caller: two days before your scheduled review. So essentially you would complete the take-home,
15:13 You: Thank you.
15:15 You: Thank you.
15:16 Caller: have a 30-minute, you know, take-home review of your code and kind of go through the core
15:17 You: Thank you.
15:19 You: Excellent.
15:23 Caller: logic. So you'll have your choice there. And then we move to the final round, which is two
15:26 You: Thank you.
15:27 You: Excellent.
15:29 Caller: interviews. It is a 45-minute product sense interview and then a 30-minute
15:32 You: Okay.
15:33 You: Okay.
15:34 You: That's
15:35 Caller: principals interview. And as we work through the interview process, you and I would set up time
15:36 You: I'm looking
15:39 You: to get in the first
15:40 Caller: to prep for these interview rounds and, you know, discuss them, answer any questions that you
15:45 Caller: might have prior to them being scheduled.
15:53 You: first one scheduled.
15:55 Caller: Yeah, absolutely. I will get your info over to the
15:57 You: Perfect. Thank you very much.
15:59 Caller: the hiring team today. And as soon as we have an update, we'll reach out with next steps accordingly.
16:08 Caller: You are very welcome, Jed. Thank you for your time today, and I hope you enjoy the rest of your day.
16:10 You: You as well. Take care, Ashley.
16:16 Caller: Thank you. Bye.