OpenAI Software Engineer Interview Experience — a Referral Bump, a Rushed KV Store Round, and a CI/CD System Design

This was for an infra engineer role on the data platform team. We chatted briefly, compared backgrounds, and it seemed like a good match. He then asked if I had any connections at OpenAI. I checked LinkedIn and found two former coworkers who worked there, so I gave him their names. I'm guessing he used them as references, because two days later I got an email moving me into the phone screen stage. The recruiter had already walked me through the interview process in our half-hour call, so everything after that moved quickly and smoothly.

The phone screen had two parts: a coding round and a system design round.

Coding: the recruiter had told me beforehand it would be an in-memory data store, and that I should know how to serialize it in Python. Since I'd already interviewed on an in-memory database question at another company, and I'd seen a similar question posted here for this company, I didn't prepare anything else. But when the interview started it turned out to be a KV store problem, which threw me off a bit. Just understanding the problem statement took a long time, and by the time I finished writing the code, the tests didn't pass and time was already up. I probably failed this round.

System Design:

• Design a multi-tenant CI/CD system that schedules and executes user-defined workflows after receiving a Git push
• The system receives information about the push through an API call from an internal service, including the repository ID and the current state of the repository
• A workflow consists of a sequence of jobs that run in order, defined in a single YAML file at a fixed location in each repository
• Users should be able to view the output and status of jobs while they're running

The engineers who interviewed me were all nice. I got the rejection email two days later.