Mock: project deep dive
45 minutes on one of your own projects, with the interviewer pushing on decisions, numbers and your personal role.
Third-party reports describe a technical project discussion in OpenAI's process, and Staff loops elsewhere commonly include one (Sources O9, third-party). Choose a project from your story bank where you made the key technical decisions.
Your partner should ask
- Draw the architecture. Why this shape and not another?
- What were the scale numbers? Which one was the real constraint?
- What broke in production? How did you find out?
- Which decision would you reverse now?
- What exactly did you do versus your team?
- If you had half the time, what would you have cut?
This rubric is ours, built from published signals (for example OpenAI's "well-designed solutions, high-quality code, optimal performance, and good test coverage"). It is not any company's scoresheet. See Sources.
Project deep dive
45:00
- Context and your role (5 min)
- Architecture walk-through (12 min)
- Interviewer drills in (20 min)
- Reflection and lessons (8 min)
Score yourself (or have a partner score you)
Past attempts
Practice
Written for this site to drill the lesson. Not reported interview questions.
- Do this mock for two different projects: one infrastructure-heavy, one customer-facing or AI.
Check yourself
What should you prepare one level deeper than the story?
Architecture, scale numbers, failures and the decisions you would reverse.
Saved in this browser.