At Source Gate Limited, I test a multi-tenant AI platform and evaluate its RAG features against expected answers and independent ground truth I build from source documents in Python. I use citation and debug traces to distinguish retrieval faults from answer-construction faults, and track groundedness, hallucinations and run-to-run variance.
Before focusing on AI evaluation, I worked in software QA at Bentley Systems, Avenir Technologies and Teralight. At Bentley Systems, I led the Framework Certification QA team and tested AutoPLANT applications; I’ve also built Playwright automation projects with TypeScript and GitHub Actions.

