Boopathi Raja

I've authored 30+ adversarial coding-agent benchmark tasks for Handshake AI's Terminal-Bench 2 / Harbor, producing specifications, reference solutions, and sealed Python, Docker, and pytest verifier suites. Four tasks cleared all 17 automated review stages at a 0/5 agent solve rate.

As an independent engineer, I deliver and maintain production systems for clients in India and Australia, including Steel Flow, Women’s Zone, The Signature, Prepli, and Nexq. For Nexq, I built a digital queue and booking platform with Stripe billing, OTP onboarding, multi-tenant security, Playwright end-to-end tests, and CI auto-deploy.

I also build and evaluate AI systems end to end—from retrieval-augmented generation and fine-tuning experiments to on-device C++ inference, agentic security tooling, and Unity ML-Agents self-play. I use evidence and rigorous test design to decide what ships.