went through the Mistral AI product manager interview process for a senior PM role. won't share the outcome but I'll share the content.
first thing to know: this is not a typical PM interview. no 'design an alarm clock' product sense questions. no 'estimate the number of piano tuners' market sizing. those are for companies that hire PMs by the dozen. Mistral has maybe a handful of PMs, they're not running a factory.
what they actually ask:
product vision and strategy: how do you think about building a product on top of a rapidly changing model? what happens to your roadmap when the underlying capability doubles? what does good product judgment look like at a foundation model company vs. an application company?
technical depth: they will ask how models work at a level that requires you to know something real. not implement-an-attention-layer level, but understand tokenization, context windows, inference cost, hallucination patterns. surface-level 'I work with AI products' isn't enough.
prioritization: given constrained compute and research bandwidth, how do you decide which product investments to make? they want to see real trade-off reasoning, not frameworks.
behavioral: describe a situation where you had to convince an engineering team to build something they thought was the wrong call. what happened? how have you handled a roadmap that got invalidated by an external event?
overall vibe: they're looking for a PM who thinks like a builder, not a coordinator. if your background is mostly slide decks and stakeholder alignment at a big company, this is probably not the right fit. they want someone who's comfortable saying 'I don't know, let's find out' and then actually finding out.
comp was not discussed until after the final round, for what it's worth.