I build software systems — full stack, with a current focus on RAG pipelines and MCP integrations.
- programming: C++, Python, TypeScript, Go, SQL
- backend: Node.js, FastAPI, PostgreSQL, Prisma
- frontend: React, Next.js, Tailwind, React Native
- ai/data: Langchain, RAG, MCP
- tools: Git, Docker, Linux, Postman, Figma
- Er. Forge
AI system that connects to your LeetCode account, diagnoses why you keep failing the same problem types, scores you across 12 skill concepts, and assigns targeted practice — with a shareable recruiter report at the end.
Next.js 14Node.js + ExpressPostgreSQLOpenRouterGoogle OAuth + JWT
LeetCode has no official API — authenticated against their internal GraphQL endpoint using session cookie + CSRF token passed by the user
- Toonie
Describe an animation in plain English, get live p5.js code running in the browser instantly — then record clips and assemble them into a single exported video on a timeline editor.
React + ViteNestJSp5.jsOpenRouterTypeScript
Dynamically executing AI-generated p5.js inside React without crashing the page required sandboxing sketch execution away from the main React tree
- WhyOpen
Chrome extension that silently classifies the intent behind every tab you open — Research, Work, Entertainment, etc. — and shows a live badge for all open tabs in the popup.
Chrome MV3Node.js + ExpressPrismaNeonGoogle Gemini
Built a two-layer cache — chrome.storage.local first, Neon DB second — so the AI is only ever called once per unique URL
Keep it simple · Make it observable · Fix what breaks
- deeper system design and backend fundamentals
- moving from demos to systems that can actually be used
Lifting weights, capturing frames, sketching ideas, and occasionally losing games I was sure I'd win.