Affirm 5-round loop - HM round tanked a clean technical process
Affirm 5-round loop - HM round tanked a clean technical process
Result: Rejected Number of Rounds: 5
Applied to a fintech role mid-level. Got the phone screen a week after, recruiter gave the impression the queue was moving fast.
Phone screen was the loan/parent company hierarchy problem. Classic fintech context. Loans associated with companies, companies with parents, walk up the tree to find topmost parent. There was 100+ lines of starter code to read before touching anything or you'll miss constraints and fail edge cases. Finished it but the interviewer was stone-faced.
Second round was fraud detection. Given customer event logs with PII data, find fraud flags. Two parts: debug existing code, then implement detection logic. Not algorithmic at all, just careful data handling. Got through both with time to spare.
System design was A/B testing. Standard experimentation platform, assignment bucketing, metrics collection. Interviewer was engaged here, we went a bit over time.
Then the HM round. The sketch business started immediately. He joined 8 minutes late. Seemed like he hadn't read my profile. Asked a mix of "how" questions. Not standard STAR format. Answered but the energy was off the whole time.
PM round was fine. Very casual.
Rejected with no feedback. HR said they moved forward with other candidates and went dark. Every technical signal was positive. The HM round broke it.
Replies (5)
Did the SD interviewer extend time past 60 minutes or was it strictly capped?
Yeah we went maybe 15 minutes over. That round actually felt positive.
The loan hierarchy problem with 100+ lines of starter code is peak Affirm. Reading comprehension test disguised as a coding interview.
The HM round killing an otherwise clean loop is such a documented Affirm thing. I've warned a few friends about this.
Filtering out qualified people because of one disengaged HM is a waste. Sorry this happened.
Did you know which team you were interviewing for going in? Some teams have a worse rep than others.
Payments infra team. Thought I was safe but apparently not.
Hi, For the PII data problem, what did the problem look like? Also, how did you approach correcting the incorrect method? For the second problem, was it about matching transactions to fraud events? I tried looking for patterns but couldn't find any clues. Thank you in advance!