AI engineer & full stack developer in Barcelona

Solid foundations and shipped work. What I reach for by default.

• LLM APIs (OpenAI · Anthropic)Handle the messy, human parts of an app that rigid rules never could.
• Prompt engineeringTurns a nice demo into something reliable enough for real customers.
• MCP (Model Context Protocol)One clean way to connect AI to real tools, instead of a fresh hack each time.
• PythonMy quick sketchpad for trying an AI idea before building it for real.
• TypeScriptCatches silly mistakes before users ever see them.
• React.jsBuild screens as small blocks I can reuse instead of repeating myself.
• Next.jsOne tool to build the whole site, so I can ship it on my own.
• Node.jsRuns the behind-the-scenes work cheaply, in the same language as the rest.
• +10 more

Comfortable building with these today, still deepening the edges.

• RAG pipelinesGrounds answers in real docs to kill hallucinations on domain questions.
• LLM tool callingLets models take actions and return structured data, not just prose.
• Generative video & image pipelinesMakes cinema-grade frames viable for a team with no VFX budget.
• CI/CD pipelinesAutomates tests and deploys so shipping stops being a manual risk.

Actively skilling up on the agentic stack, the direction I'm heading.

• LangGraphLearning it to make agent flows debuggable instead of a black box.
• GraphRAGExploring it for retrieval where relationships matter more than similarity.
• Vector databasesGetting fluent so semantic search scales past a toy dataset.
• Agent evals & guardrailsLearning to measure agents so I can trust them in production.
• Graph databasesStudying them to model connected data that SQL makes painful.
• Testing (unit & E2E)Building the habit so refactors stop breaking things silently.