The best way to understand agents is to build them. The best way to trust them is to ship them.
I'm an AI Engineer at a stealth AI startup (London — remote). I build agentic systems, RAG pipelines, and multi-agent architectures. Most of my time goes into figuring out how to make AI agents actually work in production — not just in demos.
| Project | What it does |
|---|---|
| tribev2 | NeuroLens — interactive neuroscience analysis on Meta's TRIBE v2 brain encoding model. PyTorch, CLIP, Whisper, fMRI visualization. |
| Moonsense | Multi-agent AI skincare system — specialized agents for medical diagnosis, environment analysis, product matching, and safety validation. FastAPI, Next.js, Gemini, Langfuse. |
| Project | What it does |
|---|---|
| agentscope | Alibaba's multi-agent framework. Digging into MCP, A2A protocol, and production orchestration patterns. |
| MiroFish | Swarm intelligence engine — thousands of AI agents with independent personalities forecasting future scenarios via GraphRAG. |
| Project | |
|---|---|
| LangGraph | LangChain's agent orchestration framework |
| CrewAI | Multi-agent orchestration platform |
| h2oGPT | Open source LLM deployment |
| Langflow | Visual agent builder |
| Ollama | Local LLM runtime |
| Agno | Agent framework |
| Browser Use | Browser automation for AI agents |
106 commits · 1 issues · 8 pull requests · 0 code reviews
Repositories contributed to outside my own
The thread through all of this: agents are moving from toy demos to real systems. That transition needs people who care about the engineering as much as the research — orchestration, observability, failure modes, evaluation. That's what I work on.