AI-Gov Compliance Auditor.
A full-stack platform that automates a job normally done by specialist law firms over weeks: assessing whether an AI system’s documentation actually satisfies the obligations of major AI regulations.
Upload a system specification, and the platform returns a structured, scored report across up to four regulatory frameworks at once, each with a 0 to 100 score, a risk classification, supporting evidence, identified gaps, and remediation steps.
A properly distributed system rather than a single model call: a Next.js frontend, a FastAPI backend, an asynchronous job queue and worker, two separate vector databases serving two retrieval needs, and Claude doing schema-enforced scoring grounded in retrieved regulatory text rather than memorized generalities.
Every score is forced into a strict, machine-parseable format; scoring runs under bounded concurrency to respect API rate limits; and the full test suite runs without ever calling a real external API, with every model call mocked in CI, keeping tests fast, free, and deterministic. Deployed live across two hosting platforms with a complete, tested runbook.