Full-stack engineer building AI-powered products
🤖 AI & agents
- Test-planning agent: an MCP-integrated agent that reads product specs and designs, then generates manual test cases and Playwright automation. It took test design from 4-7 hours per story to under 10 minutes, with an eval layer to keep the output consistent and auditable.
- Next.js Docs Assistant: RAG over the Next.js docs with cited answers and similarity scores. OpenAI embeddings, Supabase/pgvector, LangChain, query expansion, adaptive thresholds.
- CoverLetter.AI: a Chrome extension that reads the job posting on your screen and drafts a tailored cover letter. 50+ active users. GPT-4, Node/Express, TypeScript.
- Test-failure triage agent: reads failing test runs and works out why they failed. Java. (in progress)
🚀 Products & systems
- WishHub: an India-first wishlist and price-tracking app with affiliate links. React Native/Expo, Next.js, Supabase.
- Vercel Clone: paste a GitHub repo, get a deployed site. Three Express microservices, a Redis build queue, and S3-backed storage.
- File-Transfer: serverless peer-to-peer file sharing over WebRTC. A 1 GB test transfer took about 80 seconds, with no server storage.
| Languages | TypeScript JavaScript Java Python SQL |
| Frontend | React Next.js React Native Angular Tailwind |
| Backend | Node.js Express Spring Boot FastAPI |
| Databases | PostgreSQL Supabase MongoDB Redis MySQL |
| AI | OpenAI API Claude API RAG LangChain MCP pgvector Prompt Engineering LLM Evals |
| Testing | Playwright RestAssured Tosca Selenium TestNG |
| Cloud & DevOps | AWS Docker Jenkins CI/CD |
Full-stack and AI engineer roles: building AI-powered products end to end, from the UI to the LLM pipeline. My testing background means I ship things that hold up.