AI Lab
Building with AI.
Not just talking about it.
The AI Lab is where I ship software. Tools, agents, automations, and applications — built with the best AI APIs and frameworks available today. Every project is a real thing that runs, not a prototype that never left a Jupyter notebook.
What I build
Autonomous systems that actually work
Agents built on top of LLM APIs that handle real tasks — research, synthesis, code generation, workflow automation. Focused on reliability and production-readiness, not demos.
Software for people who build software
CLI tools, APIs, and libraries that make AI capabilities accessible and composable. Built with the same precision I'd want as a user.
Products with intelligence built in
Full applications where AI is the core feature, not a bolt-on. From idea to deployed — exploring what's possible when you build AI-first rather than AI-last.
Projects
Project 01
Details coming soon. Building in the open — this card will update as the project takes shape.
Project 02
Details coming soon. Building in the open — this card will update as the project takes shape.
Project 03
On the roadmap. Stack and scope being defined.
The stack
AI / LLM APIs
Claude (Anthropic), GPT-4o (OpenAI), Gemini
Agent frameworks
Claude Agent SDK, LangChain, custom orchestration
Languages
TypeScript, Python
Frontend
Next.js, React, Tailwind CSS
Backend & APIs
FastAPI, Node.js, Vercel Edge Functions
Data & storage
PostgreSQL, Supabase, Pinecone (vector)
Infrastructure
Vercel, GitHub Actions, Docker
Observability
Langfuse, custom evals
How I build
The best way to understand what AI can do in 2026 is to build something real with it. Reading papers and watching demos tells you what others found. Shipping tells you what you find.
Demos are easy. Reliable systems are hard. Every project is built with observability, error handling, and real users in mind — even when it's a side project at 11pm.
AI systems built as modular, composable components age far better than tightly coupled pipelines. The tools and APIs will change. Good architecture shouldn't have to.
Have a problem worth solving?
If you're looking for AI software built with architectural discipline — not a no-code wrapper — let's talk.