Fin's own tips for this round:
- Choose a project where you were deeply involved as a hands-on IC
- Prefer something delivered in production as part of an ML product, over pure research
- Relative uplift without absolute numbers is fine if the numbers are proprietary
- If something didn't work, say so. Fin values critical thinking and honest reflection
- Keep it conversational
Staff/Senior signal to aim for: cross-team or org-level scope, ambiguity you handled, defensible tradeoffs, and quantified impact, not just technical depth.