selected work
case studies & tradeoffs.
Each link goes to a case study with architecture notes, decisions, and what I'd do differently.
QA Platform for AI-Drafted SDLCs
A multi-app QA platform that survives AI-speed shipping. HTTP and API-based invariant tests guard system-wide rules. Self-healing browser tests absorb UI drift. A backend orchestrator runs everything with queueing, test-account locking, manifest-driven coverage, and Jenkins integration. A Next.js 15 dashboard surfaces the operating model.
Manual QA Agent
An agentic tester that behaves like a senior QA engineer. Point it at a commit, ticket, or spec; it derives scenarios, executes across blackbox / DB / API layers, and returns a findings report with evidence. Pairs with invariant tests: always-on + on-demand. A staged knowledge base keeps the corpus trustworthy over time.
Lumineltek: RAG-as-a-Service
Multi-tenant SaaS platform for AI-powered document processing and semantic search. 4-phase ingestion pipeline with checkpointed pg-boss workers, hybrid retrieval with swappable search providers (pgvector or Elasticsearch), 40+ permission domains, metered billing via Lemon Squeezy, and a public API for external consumers. Built end-to-end.