What it is
My portfolio doubles as a full-stack AI platform. Visitors can chat with an agent grounded in my experience, or talk to a 3D avatar of me that answers out loud.
How it works
- RAG chatbot: portfolio content is embedded into Pinecone and retrieved per question.
- Streaming agents: the OpenAI Agents SDK runs tool-calling agents whose replies stream token-by-token to the browser over Socket.IO.
- Talking avatar: a 3D avatar (TalkingHead + three.js) speaks each sentence as soon as it arrives, with lip-sync from in-browser neural TTS on WebGPU.
- Self-hosted: a custom Node server runs Next.js, Socket.IO and scheduled jobs in Docker, with a free LLM gateway behind it.
- Admin suite: CMS, kanban boards with AI standups, and a finance module with its own agent.