VendorVerdict
Lead AI Developer (contract, UK) · May 2026 — present
An AI platform that gives CISOs evidence-backed due diligence before they trust a security vendor with millions — architected so unverified evidence can never change a verdict.
Before signing a major security contract, someone has to prove the vendor itself is safe — and that evidence is scattered across government vulnerability feeds, financial filings and enforcement records. I built the system that unifies six authoritative sources, scores every vendor across product-security, financial and regulatory risk, and generates the decision brief. AI runs inside the pipeline, but the engineering is the point: every conclusion is traceable to its source, factual claims are kept separate from estimates, and the whole thing is built to fail safe.
- AI-assisted evidence processing designed to fail safe — unverified input can never influence a verdict
- Six authoritative feeds unified and continuously scored: NIST NVD, CISA KEV, SEC EDGAR, Companies House, vendor PSIRT, ICO enforcement
- Versioned, explainable methodology — every material conclusion reproducible and defensible after the fact
- Rode out a real upstream-source outage in production with zero intervention; 8,073-request soak test, zero errors; full OWASP A01–A10 audit
8,073
requests, zero errors
6
authoritative sources
Fail-safe
by design
Python · FastAPI · LLM pipelines · Prompt-injection defense · Evidence integrity · OWASP · PostgreSQL · Next.js