AI coding interview practice

AI-assisted coding rounds hand you a codebase you've never seen and an AI assistant. You fix a bug, build a feature and speed something up, and the interviewer watches how you use the AI. These 3 free projects let you rehearse exactly that on your own machine.

• Inventory reservations

  A checkout service that holds stock while shoppers pay. Fix a double-counting bug, add all-or-nothing cart reservations, then make stock checks fast during a sale.

• Log insights

  An on-call dashboard that reads gateway access logs. Fix a parsing bug, add an error-rate panel, then make refreshes fast with a day of traffic loaded.

• Rate limiter

  A per-key token-bucket rate limiter for an API gateway. Fix a refill bug, add per-customer limits, then make it survive 200,000 keys.

How to practise

• Set a 60-minute timer: about 10 minutes for phase 1, 25 for phase 2, 20 for phase 3 and 5 to review.
• Use an AI chat that can read the project files (or paste them in), but type every change yourself. Many AI-enabled rounds work this way: the AI explains and suggests, you edit.
• Say your plan out loud before you prompt, and check every AI answer against the code and the tests before you trust it.
• Run only the phase you’re on, then the whole suite, so you know you didn’t break what already worked.

Is this the real format?

It follows what candidates describe: a multi-file project, phases that build on each other, and tests that decide whether you're done. Details differ by company. See which companies allow AI and the candidate reports for each one. Every project here is original code, not taken from any company's interview.