Each scenario gives you a broken production service — the kind an FDE inherits on day one. You diagnose the bugs, fix the code, and run the service. Programmatic checks and an LLM judge score your work automatically. Results in under a minute.
Every submission is graded automatically across three areas. No self-reported skills — only evidence from your running code.
Config checks
~40 ptsProgrammatic verification that you changed the right values. Did you identify and fix the silent bugs in the configuration? Each check is binary — pass or fail.
Live query tests
~40 ptsEnd-to-end functional tests run directly against your running service. Your fix has to actually work — not just look right. Each query is scored on the real output.
Code quality
20 ptsAn LLM judge reviews your git diff, logs, and test outputs to score your investigation approach, solution quality, and engineering rigour — not just whether the tests pass.
Each scenario is scored out of 100. Where you land tells you exactly where to focus next.
Developing
Config fixes missed or tests failing — revisit the root cause
Deployment-Ready
Fixes applied but some queries still returning wrong answers
Battle-Tested
All checks passed with solid investigation approach
Principal FDE
Clean fix, all tests green, LLM judge rated your approach highly