Engineer at TestMu AI (formerly LambdaTest). I build the layers most people import.
- nanodriver — a browser driver written from scratch: an RFC 6455 WebSocket client on a raw TCP socket, and a Chrome DevTools Protocol client on top of it. Python standard library only, no Selenium, no Playwright. Built to measure what actually makes browser automation slow. In progress.
- webguardx — full-page web audit on Playwright: HTTP status, WCAG 2.x via axe-core, broken links, performance, console errors, with 5 built-in auth strategies and CI-ready exit codes.
npm install webguardx - gatecrasher — a benchmark that measures how often browser agents obey instructions they were never given. Frozen pages, attack success rate over N runs, usable as a CI regression gate. Early.
- Redis in C — an in-memory key-value server with
select()-based multiplexing,SET/GET/DEL, and a custom CLI client. - ByteScribe — full-stack interview-prep platform (FastAPI, React, PostgreSQL, Redis) built to exercise auth, real-time features, and AI integration in one codebase.
Also contributing to Travsr, a graph-native code intelligence daemon that serves AI agents structural context over MCP (Rust).
I own the test infrastructure behind KaneAI, TestMu AI's GenAI-native testing agent:
- three production frameworks from scratch: an API suite (192 endpoints / 32 modules), a YAML-driven E2E framework (650+ scenarios), and a Playwright UI suite
- EvalSet pipelines that score LLM output across authoring, code-generation, and execution
- an agentic workflow that reads PRDs and GitHub PRs and produces reviewed test cases, cutting manual test-creation effort by ~90%
Python · TypeScript · C · Go · Playwright · pytest · FastAPI · Docker · GitHub Actions