Turn the work into a job · 13 / 15

Prepare to explain, debug and design

AI Engineer Path15 min guide

Practice the work you are likely to do in the role. Ask the recruiter what the interview covers. Some teams emphasize coding and algorithms; others focus on systems, customer scenarios or project review. The exercises below are my preparation rubric, not a description of every employer’s process.

Four practice sessions

  1. 45-minute coding: parse a file, deduplicate records, call a fake API, handle an error, and test the result. Explain complexity and edge cases.
  2. 45-minute debugging: diagnose why a retrieved answer cites the wrong document. Separate ingestion, retrieval, prompting and generation failures.
  3. 45-minute design: design a document assistant for 100 internal users. Clarify data access, freshness, success criteria, budget, failure modes and rollout.
  4. 30-minute project defense: give your demo, explain your baseline, show a failure and describe what you would change with another week.

Questions you should answer in plain language

  • Why did you use retrieval instead of a larger prompt or fine-tuning? What experiment would change your mind?
  • How do you know your evaluation set resembles real use? What did you hold out?
  • How would you prevent one user from retrieving another user’s documents?
  • What happens if the provider times out, the model changes or a document contains malicious instructions?
  • What does one successful task cost? Where is most latency spent?
  • What did AI assistance write, and which decisions, tests and reviews did you personally own?

Readiness rubric

Score each area 0 = cannot do it yet, 1 = can do it with help, 2 = can do it and explain it: coding/debugging, API/data flow, LLM concepts, evaluation, deployment, and communication. Aim for no zeros before a technical interview. This is a study aid, not a predictor of an offer.

Keep ordinary software interview practice in your schedule, especially for backend and full-stack openings. Respect each employer’s rules about AI tools during applications and interviews. Practice both with tools and without them so you can show your own understanding.

Move on when

You can explain your architecture without reading notes, solve a modest coding task, and discuss a real failure with the evidence that led to your fix.

Your lesson resources

Download these files to follow along and put the lesson into practice.

Saved in this browser.