Hyperloop UPV GUI
California, USA
I contributed to the team's graphical interface for monitoring a Hyperloop prototype, with views for motors, brakes, batteries, and raw data.
Software engineer · AI infrastructure
Hi, I'm Carlos. I build AI systems in restricted environments.
I build and operate on-premise AI infrastructure and LLM systems where security and reliability matter. I bring more than six years of production engineering experience across automotive, energy, and industrial technology.
See what I'm working on
What I'm working on
Production AI, from the infrastructure underneath it to the tools people use every day.
At Rohde & Schwarz, I serve Mistral, Qwen, and GLM models with vLLM on Kubernetes, OpenShift, and ArgoCD in a high-security environment.
- Designed ways to adapt on-premise model execution to demand using Run:ai, vLLM, and DCGM GPU metrics.
- Built Grafana dashboards that make AI infrastructure health and model usage clear to engineers and management.
- Built MCP servers and Skills that extend what AI agents can do across the company.
- Integrated Claude Code, Codex, and OpenCode into enterprise development workflows under strict on-premise requirements.
- Took ownership of production systems and helped onboard colleagues beyond my defined role.
I also help people put these tools to work. I've delivered agentic coding sessions for engineers and management and driven AI upskilling for more than 50 engineers, including our Spain location. My R&D interests include AI audio, music generation with Stability AI models, and TTS workflows with Parakeet v3 and Nemotron 3.5.
Earlier team projects
Some of the software work that shaped how I build today.
Radiotherapy plan review
Helsinki, Finland
With the Makers UPV team, I worked on a web app for Varian's HackJunction challenge that helped oncologists review treatment plans and compare dose information.
Writing and teaching
I share what I'm learning about AI tools and the way we work with them.
LinkedIn post
Inside our Agentic Coding School
What I learned while helping colleagues move from simple prompts to hands-on agentic workflows.
Background and tools
Six-plus years of production engineering across Volkswagen, Techem, and Rohde & Schwarz.
My background spans full stack development and production operations. I studied at Universitat Politècnica de València and now focus on making AI systems useful inside real enterprise constraints.
Tools I work with
Kubernetes, OpenShift, ArgoCD, Docker, CI/CD, Run:ai, vLLM, DCGM, Ollama, LiteLLM, MCP, RAG, Mistral, Qwen, GLM, Java, Spring Boot, TypeScript, React, Next.js, AWS, and Azure.