Personal / side project
FlawFerret2
AI-powered QA orchestration platform: a Chrome extension captures browser testing context (DOM, screenshot, console errors, network info), which is turned into a queued job that an AI coding agent (OpenAI Codex) implements against a target repository, with Playwright validating the change and a draft GitHub pull request opened for human review. Humans remain the approval gate at every AI-driven step.
Pipeline
Browser extension -> Next.js web dashboard -> Fastify/Zod API -> Prisma/PostgreSQL (Neon) -> ferret-runner worker -> Codex (AI coding agent) -> Playwright validation -> GitHub draft pull request
My role
- Sole architect and builder — product vision, system architecture, database schema, subagent pipeline design, and safety-gating model
- Designed the human-approval gating model that scopes what an AI coding agent is allowed to do autonomously vs. what requires explicit human sign-off
Stack
Scope & status
Prototype/early-stage personal project, not production software or shipped-at-scale: as of the current milestone, repository checkout, Codex invocation, Playwright validation, and GitHub PR automation exist as scaffolded code but are gated behind default-disabled flags and manual approval steps rather than running unattended in production. Represent as hands-on personal/side-project experience with agentic AI engineering, MCP tooling, and AI-assisted SDLC design — not as professional/paid AI-ML employment or production deployment at scale.