interviewing at Visa for a senior data scientist role in 5 weeks. i've found a decent amount of info on the SWE loop but almost nothing on the data side.
if you've interviewed at Visa recently (last 6-12 months) for any data, analytics, or data science role, can you drop what you saw? specifically curious about: SQL complexity (basic joins vs. window functions vs. full optimization problems) whether there's a take-home or live case how heavy the ML component is vs. analytics anything about the fraud/risk team specifically
any SWE loop data is also useful for comparison. thanks in advance.
4 replies
analyst_ana
i interviewed for a data analyst role (not DS, slightly different) about 8 months ago. the SQL round was medium-hard. definitely window functions, a running total problem, and one question about detecting duplicate transactions in a table. nothing i'd call brutal but you needed to be comfortable with CTEs.
ml_mike
did a loop for an ML role on the fraud team 9 months ago. got a take-home: build a classifier on a synthetic transaction dataset, flag anomalies, explain your feature choices. 3 days. then a 30-min debrief where they pushed on why i picked the threshold i did. the debrief was harder than the take-home. they wanted me to defend the business tradeoff between precision and recall, not just the technical answer.
ds_dmitri
precision/recall tradeoff in a fraud context is basically always about false negatives vs. false positives, right. false negative = missed fraud, false positive = declined legit transaction and angry customer. i can prep that angle. was the debrief with a hiring manager or an IC?
ml_mike
IC, senior ML engineer. no manager in the room for that round. manager came in the behavioral round separately.