have a data science loop at Affirm coming up in 3 weeks and I want actual recent data. the role is on the credit risk modeling team.
from what I've gathered so far: SQL round is expected, probably one ML conceptual round, maybe a case or product sense component. but I'm going off stuff from 2023 and processes change.
if anyone has done a DS or analytics interview at Affirm in the last year, drop your loop structure here. I'm especially curious: how deep does the stats/probability go? did they test ML implementation or more conceptual? was there a take-home component? any surprises?
thanks in advance.
4 replies
analyst_ana
did a data analyst loop there about 8 months ago. for analyst it was: SQL screen (window functions, aggregation, one tricky self-join question), then a take-home case involving loan default rates with a CSV, then an onsite with 3 rounds including one where they literally just walked through my take-home and asked "why did you choose this approach." that last part was harder than expected.
ds_dmitri
the take-home walkthrough sounds like the real test honestly. good to know they scrutinize the decisions not just the output.
ml_mike
credit risk adjacent: they will almost certainly ask you to explain a model you'd use for predicting default probability and why. and then they'll poke at calibration, at feature leakage risks, at how you'd handle class imbalance. less about writing code, more about whether you can explain the tradeoffs like you've actually shipped one.
alex_design
worth noting that Affirm's credit team has been through some headcount changes in the past year. their model risk appetite has shifted post-profitability-push. I'd spend some time reading their investor letters before the final round, the questions get more interesting if you can reference their actual business context.