The New York Times · Primly Community

The New York Times senior / L5 system design interview, what to expect and how i prepared

hardware_hugo · 5 replies

went through the NYT system design round for a senior-level position in late 2025, joining the platform team. going to write up what i saw because it's different from what i expected going in.

the prompt they gave me: design a content delivery system for a major news event where traffic spikes 50-100x from baseline in under 5 minutes. think: election night, breaking news, that kind of thing.

right away you can see why: the NYT actually deals with this. election night 2024 was one of their biggest traffic events ever. so this wasn't a generic 'design twitter' prompt, it was something directly tied to their engineering reality.

what they cared about: caching strategy at multiple layers (CDN, app layer, DB read replicas). they pushed back on every assumption. how do you handle cache invalidation when an article updates mid-spike? good question. database write bottlenecks: they wanted me to go deep on read/write separation and sharding decisions. graceful degradation: what does the site do when the origin can't keep up? serve stale? queue? i spent a lot of time here. they asked about how i'd instrument this. observability wasn't an afterthought for them.

what they seemed less focused on: pure ML recommendation systems, microservice decomposition for its own sake, any fancy distributed consensus stuff. they wanted boring reliable infra that survives chaos.

format: 60 minutes. one interviewer, asked probing follow-ups constantly. no whiteboard drawing tool -- just verbal explanation + i sketched rough ascii in the shared doc. that was fine.

how to prep: honestly, study how CDNs work (Fastly specifically, they use it), read a few public NYT engineering blog posts, and do a couple system design mocks where you practice the graceful degradation angle. 'what fails first and how do you not let it cascade' is a good mental model for their prompts.

the bar felt like: can you reason clearly under pressure about a real system, make defensible tradeoffs, and talk through failure modes without being prompted. if you can do that, you're fine.

5 replies

de_derek

the content delivery under traffic spike prompt is interesting. very specific to news/media infra. i interviewed at a streaming company and got a similar 'peak event' angle. good reminder to think about the company's actual scale problems, not just generic 'design instagram' stuff.

ml_mike

did they ask anything about the recommendation engine or article ranking? or was it purely infra? NYT has a substantial ML team but curious if it bleeds into SWE loops.

staff_steph

not in my loop -- i was going for a platform/backend role so it was purely infra-flavored. if you're targeting the ML/recsys team it's probably a totally different process. might be worth asking the recruiter which team the role sits in.

sec_sasha

no one ever asks about auth/access control in system design rounds and it drives me insane. did they bring up anything around authenticated vs. anonymous user caching? that's a real pain point for news sites.

backend_bekah

the cache invalidation on a live-updating article question is the exact kind of thing that sounds simple until you think about it for 30 seconds. real edge case, good sign they're asking it.