Sharing my Snap TPM interview from a while back.
Round 1 with the HM: A quick 30-minute chat with the hiring manager, mainly talking through two projects I'd worked on before and digging into the details. They also gave me a heads-up on what to focus on in each of the following rounds.
Round 2, with the HM again: This round was more free-form — the HM asked a wide range of questions covering statistics, infra, and data science. The ones I specifically remember:
- What is Recall
- What is ROC, AUC
- What is P-Value
- How to do RCA if a service's SLA drops, and how to improve that
- Tell me a time you worked with a difficult person
- Tell me something counterintuitive you learned
- What is hypothesis testing? What is Type 1 and Type 2 error?
Round 3, with a Senior TPM: This one tested coding (SQL + Python), easy-level questions. I'd actually expected DFS/BFS-type problems and spent time grinding Leetcode for that, but it turned out to just be basic data-manipulation questions. 12 questions in 30 minutes — you weren't expected to finish all of them, but you had to be able to explain your reasoning.
After that I moved on to the onsite. The onsite had four interviews total.
Round 4 — 30-minute presentation: present a project you led before. About 30-something Snap TPMs across various levels showed up to listen, then asked questions afterward.
Round 5 leaned more toward daily operations, interviewer was a TPM from another team:
- Do you have experience with cost-saving?
- How do you evaluate a project's ROI?
- How do you make teams accountable?
- Have you run into a situation where SLA dropped, and how did you resolve it?
Round 6 leaned more toward product sense, interviewer was an eng director. Mostly we talked about the product I currently own — what the metrics are, whether I've run into a situation where the numbers looked good but my boss wasn't satisfied and how I led the team through that. Basically just a deep dive into what I'm currently working on.
Round 7 leaned toward SW-related knowledge:
- Tell me about an engineering project you were hands-on with, end to end.
- What does the ML Lifecycle consist of?
- Suppose Snap wanted to build a content-recommendation service — walk through what it would need, applying the ML Lifecycle you just described.
The whole process took about two months start to finish, and this was the last interview I had for a while. The day after, they told me I didn't get it. I was really upset hearing that at the time, because I felt like it had gone well and I'd put a lot of effort into preparing. Looking back, maybe my communication skills still have room to improve...
Discussion
Loading comments…