I build AI systems at the intersection of research and production. Hard problems with people who care about shipping things that work.
Currently leading AI at Zylon, building private AI infrastructure for enterprise customers in regulated industries.
0k+
GitHub Stars
open source
0+
Years GenAI
at Zylon
0+
Years Engineering
Software Engineering
Product AI Engineer
May 2024 – Present
Lead AI product development end to end: setting technical direction, shipping features, and turning a greenfield product into a platform used by enterprise customers in regulated industries.
PrivateGPT
Main contributor. 57k+ ⭐ open-source backbone of the platform.
Model serving
Triton + vLLM on Kubernetes, fully on-premises deployments
Sandboxed execution
Secure runtime that lets LLMs perform real actions safely
Agentic systems
Tool calling, MCP, RAG. LangChain, LlamaIndex, custom framework.
Software Engineer
Sep. 2020 – May 2024
Software Engineer Intern
Sep. 2019 – Apr. 2020
Deep in the OSS AI ecosystem. Main collaborator on PrivateGPT, contributor to LlamaIndex and Ollama. Most of production AI runs on code someone shared for free.
Deploying models on your own hardware: vLLM and Triton on Kubernetes with GPU Operator, serving LLMs, embeddings, rerankers, and OCR fully on-premises, including air-gapped environments.
Agents that do real work: tool calling, MCP servers, RAG over large document corpora, and sandboxed code execution. Building the infrastructure that lets LLMs act, not just respond.
Agent memory and long-horizon reasoning: how AI systems build, retain, and use knowledge across sessions. Evaluation frameworks, structured reasoning, and where GenAI meets classical software.
zylon-ai/private-gpt
0k+
GitHub stars
Main collaborator — 57k+ GitHub stars. The open-source backbone of Zylon's AI platform. Shipped core features: code execution, MCP integration, and multi-provider inference.
run-llama/llama_index
ollama/ollama
Contributed during early adoption while building private AI infra at Zylon.
source I've worked with closely
ecosystem tools
University of Oviedo · sep. 2026
Topic: memory systems for LLM-based agents.
The question I care about: how do agents build, retain, and use knowledge across long interactions? This connects directly to production systems I work on — where agents need to plan, act, and recover across complex workflows without losing context.