AI Chat
Agentic AI — Portfolio Project
Meet JB Proxie. I built this custom chat agent to help visitors like yourself learn more about me in an engaging and interactive way — and it's a live demonstration of my expertise in designing and deploying production AI systems.
For the technology stack, JB Proxie runs entirely on my own hardware: a personal NVIDIA DGX Spark supercomputer, using Qwen3.6 35B via Ollama for local inference, served through a custom FastAPI backend with a Cloudflare and Tailscale tunnel for secure public access. The avatar visuals were generated using Google Gemini, and refined to create a consistent character identity across the interface.
To establish its knowledge base, I built a RAG (retrieval-augmented generation) system over my published research papers and portfolio, so JB Proxie can answer questions accurately rather than hallucinate. The full knowledge base includes:
- My resume, including education, work experience, and relevant skills
- RAG over my publications and research papers
- My philosophy for designing systems that serve their users first and foremost
- Some fun personal facts about my interests and hobbies
- Governance guardrails to keep responses accurate, on-topic, and NDA-compliant
JB Proxie can learn too. It can identify gaps in its knowledge base and message me through Discord to discuss potential updates — taking initiative in expanding what it knows, but only with my express control and oversight. This human-in-the-loop design is intentional: it reflects the same principle that runs through all of my work.
If you have any questions or would like to discuss a potential project or work engagement, please reach out to hi@jeffxr.com. Thank you!
