Spotify · Primly Community

Did the Spotify ML loop for recsys -- here's what actually mattered

staff_steph · 4 replies

Just finished my loop for a Senior ML Engineer role on their recommendations team. Sharing what I found useful.

5 rounds total: Recruiter screen (30 min, pure logistics) Tech screen with the hiring manager -- mostly ML concepts, some Python, light on leetcode ML system design (90 min, the hardest round by far) Coding + ML problem-solving (a mix, not pure DSA) Behavioral + values (45 min, way more probing than I expected)

The ML system design round was serious. They gave me a prompt that was basically 'design a personalized playlist ranking system at Spotify scale.' You need to think through: what signals, how you'd handle cold start, latency constraints, feedback loops, A/B testing strategy. They want to see you actually think about the product tradeoffs, not just rattle off ML architecture.

The behavioral round had three full STAR questions. 'Tell me about a time you disagreed with a stakeholder about a model's direction' was one. They pushed on specifics, what did YOU do, not what the team did.

Got the offer. Total loop was about 5 weeks from first contact. Turnaround on the offer itself was fast, under a week after final round.

4 replies

analyst_ana

the 'cold start' angle is interesting -- did they expect you to already know Spotify's actual architecture or was this more generic 'here's a streaming music platform, design it'?

ml_mike

generic. they weren't testing spotify domain knowledge, more your ML system design instincts. i mentioned bandit algorithms for cold start and they lit up a little. signal that they use something like that.

sre_sol

five rounds is pretty reasonable for ML actually. i've seen 8 at places that then ghost you for 6 weeks. how was the recruiter communication throughout?

corp_refugee

the behavioral push on specifics is very spotify. every former spotify person i know says they care a LOT about individual ownership, not just team outcomes. worth prepping your STAR stories to be first-person before you get there.