Discord Senior+ Backend Engineer Interview Experience — A Redis Curveball on the Etcd Round

Sharing a Discord interview experience — I interviewed for a senior backend position. I really like Discord, I use it pretty much every day for gaming, so I do have a soft spot for this company.

Overall impression

The interviewers had mostly been at the company for 6-10+ years, very professional, technically strong, and the interview experience was great — it basically felt like solving a problem together with a colleague. Onsite also came with a free month of Nitro. The onsite interviews were conducted over Discord itself, which worked pretty well and felt like a fresh experience.

After talking to them, my sense is that internal promotion is probably hard — the interviewers were mostly still at the title they joined with, no change in 5-10 years, with a few exceptions who "rocketed." I checked the pay bands and they're not very competitive for the Bay Area. So even if I pass, I might not end up going.

Phone screen

The classic design a chat server question, written with asyncio — exactly the same problem as before. I wrote part 1 and part 2, the interviewer's feedback was good and I passed, no legendary part 3 or part 4. I got the sense this round is mainly about communication.

Manager round

Basically behavioral, checking whether my background is a match.

Onsite — 5 rounds

One round was the legendary etcd leader-follower problem. The interviewer didn't come right out and say to use etcd — he just posed the problem and said I could use whatever I wanted. I felt like saying etcd outright would be too obvious, even though I'd practiced that a lot, so I said I'd try Redis instead. I hadn't prepped Redis at all beforehand, and at first I thought I'd shot myself in the foot, but with the interviewer's help I actually managed to finish it. This round felt great — I felt like I challenged myself and learned a lot, though thinking back now it was a little reckless. What surprised me was that the interviewer knew the Redis docs inside and out — genuinely impressive.

One project deep-dive round and one behavioral round, both pretty standard questions.

One debug round, also pretty interesting. Basically it's about how you handle an incident. It starts with an alert saying CPU suddenly spiked, and then you can ask him anything — he'll draw whatever metrics and logs you want — and you need to find the root cause and fix the problem. Very open-ended. This round felt good too, the interviewer gave a few small hints along the way. My approach was to mitigate first and find the root cause after.

Last round was system design on Slack. I'd figured it'd be the eternal game-matching question, and this round's interviewer was also the most senior of the bunch. I got challenged a lot in this one — index choices, database choices, exactly how the fanout worked, all asked in detail. My brain kind of blanked out in the middle, I didn't answer well, and when the interviewer gave a hint I didn't quite get it and even pushed back a little, ha. This round is probably a bad review.

Hope this helps people interviewing later.