The OA was an hour long with two questions, pretty simple. Three days after that I got an interview invite.
Three interviews back to back, three hours total, all with very senior people interviewing.
The first interviewer worked on software/systems and had me write a problem, roughly medium difficulty. Org chart → Tree. Top-down, level by level → BFS. "Show the entire layer first" → Queue.
The second interviewer asked a SQL question, then asked about A/B testing — what to do if you can't run an A/B test (a few different ways to do counterfactual estimation), confusion matrix, Type I/II error, given a scenario how would you judge which way to evaluate a model, how to think about SHAP analysis, describe how you'd build a machine learning project end to end, and given a business scenario how would you design a testing approach.
The third interviewer was pretty relaxed about it — we just chatted for 15 minutes, they asked what I was doing before my PhD. The other 30-plus minutes were all about a business scenario they're currently working on, asking me how I'd judge various things, testing business sense — things like what metrics you'd use to evaluate whether a product is good, how you'd test it, and if a core metric suddenly drops on some day, what could be the reason, that kind of thing.
Every interviewer was really kind, very chill, and took the interview seriously. At the end of each one, during the Q&A, they'd also give a lot of thoughtful advice. If day-to-day work there is really all this kind of full-stack ML/data science design work, the overall difficulty isn't that high — the LeetCode bar is genuinely low. For most of the coding questions I didn't even finish, I just sketched out a rough approach and explained my thinking, and they let it go. Probably because it's not a SWE role.
Not sure if DS is a good path going forward, or what Microsoft's outlook looks like. I'm a bit torn on whether to accept — I have other offers and I'm still interviewing elsewhere.
Discussion
Loading comments…