Engineer · Cloud · AI · Software
I design, automate, and operate real systems end-to-end — from serverless applications to self-hosted AI infrastructure. I care about what works in production, not just in a demo.
I'm Jesús Dones Estrada — an engineer who designs, builds, and operates production systems across cloud, AI, and software. I turn ideas into things that run reliably, and I go deep enough to own the whole stack: architecture, automation, and the operations that keep it healthy.
I've built a self-hosted multi-model LLM inference platform from bare hardware to production workload, and a hierarchical multi-agent AI system with its own memory and tooling. I care about systems that are automated, observable, and honest about their limits.
Get in touchPersonal R&D and client work. The strongest is the infrastructure nobody asked me to build.
A dedicated GPU inference server, bare hardware to production workload. Runs quantized LLMs locally — a 27B reasoning/coding model and a 35B vision-capable model — with hot-swap routing via llama-swap. Tuned dual-GPU tensor distribution on ik_llama.cpp, diagnosed utilization imbalance, and caught a precision setting corrupting mixture-of-experts activations.
A hierarchical multi-agent LLM system: a routing coordinator dispatching to domain specialists, with reusable skill modules, a knowledge-base memory layer, and full architecture documentation.
Fully serverless retrieval-augmented chatbot answering questions about my background, on AWS Lambda, API Gateway & DynamoDB with LlamaIndex retrieval orchestration.
Read the docs →A serverless application on AWS demonstrating an end-to-end managed stack — API to data store — with IAM-scoped access.
View project →Self-hosted private cloud for a small business — containerized, database-backed, with Cloudflare tunnels for zero-trust access restricted to authorized devices.
View project →A three-VLAN network isolating AI/compute, wireless mesh, and management traffic, with Tailscale providing authenticated remote access without exposed ports.
Cloud, DevOps, or AI infrastructure. Have a question or a role in mind? I'll get back to you quickly.