Your goal: hold one level across two rounds, the way a loop is decided.
How this works (about 45 minutes)
- Round 1, product (about 21 minutes): a design question, a curveball and an A/B sheet.
- Round 2, ML and policy (about 24 minutes): design a detector end to end, read its numbers and choose a threshold. Ask Felix for any facts you need; they are made up for the exercise.
- Felix tells you when he switches rounds. This is a practice room in the style of a Meta loop, built from what candidates report, not the company's real process.
- If you want to sketch, ask Felix for the notepad and tell him when you have written something.
- No hints and no scores during the call. Your report comes after.
What scores well
- A user, a goal and a metric with a guardrail before any model
- Where an LLM or embeddings help (item understanding, cold start) and where simple signals win
- A policy line that is written down, and attention to the people a detector wrongly affects
- Reading an experiment or a detector sheet without being fooled by the headline
- A launch call with a stage, a stop number and what would change it
Tips
- Take the Gen AI coaches for evaluation, responsible AI and shipping first.
- Say who is hurt by each choice.