1. Configure the work
Start from a template, then add your systems, constraints, and review rubric.
Evidence-first technical interviews for the AI era
See how candidates work when AI is part of the job.
Give candidates AI, real engineering context, and room to verify their decisions. DevEval turns the work into evidence your team can inspect, not a black-box ranking.
Work, not job titles
Configure the engineering context and success criteria.
AI access is a signal
Measure direction, Discernment, verification, and calibrated trust.
Decisions stay human
Review artifacts, replay, and score rationale before deciding.
Product proof
From work brief to evidence in three views.
Click through the same core loop a hiring team uses. No feature tour required.
Backend reliability screen
Production incident
Fix duplicate payment processing without hiding failures
A retrying worker occasionally processes the same event twice. Use the provided service, logs, tests, and AI assistant to propose and verify a safe fix.
Repository
Node.js service + tests
Time box
75 minutes
AI access
Enabled and recorded
Expected output
Patch + decision note
Signals captured
- Problem framing
- AI direction
- Discernment
- Executable verification
- Shipping judgment
AI belongs in the interview
Access is not the compromise. Uncritical use is the risk.
Modern engineers use AI. The useful signal is whether they can direct it, question it, verify it, and take responsibility for what ships.
- Direct AI with clear constraints
- Catch incorrect or incomplete output
- Challenge assumptions and tradeoffs
- Verify generated work with evidence
- Decide what is safe to ship
2. Observe the process
Capture code, prompts, tests, AI critique, verification, and decision notes.
3. Review the evidence
Use the evidence packet and focused live follow-up to make the hiring call.
See it in your hiring context