AI Solutions Engineer & Lead Full-Stack Developer · Valencia, Spain
I build LLM-powered products end to end — architecture, Python backend, TypeScript frontend, model integration, deployment and operation. No handoffs, no "someone else does the infra".
Currently at Ángela Impact Economy (Data Intelligence Hub), where I lead full-stack development of internal AI tooling, including a system in daily production use with asynchronous processing and automated alerting.
Python and full-stack since 2020, LLM-based systems since 2021, professional delivery since September 2023. I came into engineering from political science and public policy analysis, which is why most of my work ends up being about the same thing: making a model's output verifiable.
| Agent Studio | Self-hosted workspace for building and running AI agents. Multi-provider chat (OpenRouter, DeepSeek, ChatGPT/Codex, local llama.cpp), MCP servers with per-call human approval and encrypted credentials, Claude-style reusable skills, Model Council (parallel models compared, then synthesised), and your own PC paired as an agent backend for real command and file execution. Hand-rolled SSE that survives client disconnect; AES-256-GCM at rest, tiered command safety, SSRF containment. TypeScript React Express SQLite MCP |
| Explainer | Full-stack app for studying academic papers with AI. Four-agent pipeline (segmentation, explanation, annotated walkthrough, resource mapping), SSE streaming progress, per-run provider selection across Gemini / OpenRouter / DeepSeek, per-project token and cost tracking, BYOK with encrypted key storage. Python FastAPI React Supabase |
| MapMyLearn — live | AI curriculum designer: generates structured, modular courses on any topic. Deployed and running on its own domain. Python React |
| Prompter | Open-source workspace for building and refining Gemini prompts — interactive multiple-choice refinement, multimodal image prompts, token preflight, explicit context caching. Next.js TypeScript shadcn/ui |
| PreceptorIA | Clinical assistant for healthcare staff: IV compatibility checks, dose calculation, interaction analysis and protocol lookup against official sources. React 19 TypeScript Gemini |
Nine-plus full-stack applications designed, built and deployed independently with an AI-assisted workflow.
Languages — Python, TypeScript / JavaScript, SQL
Backend — FastAPI, SQLAlchemy, Flask, Celery, Express · async processing, REST API design
Frontend — React, Next.js, Angular, Vite
AI engineering — Gemini, OpenAI, Anthropic and OpenRouter APIs · RAG and embeddings · LangChain, LangGraph · MCP · agent loops and tool use · structured outputs, guardrails and output validation · streaming · evaluation
Cloud & infra — GCP (Cloud Run, BigQuery, Pub/Sub, Secret Manager), AWS (Bedrock, Lambda, S3, ECS), Docker, GitHub Actions, Vercel
Testing — pytest, Vitest, Playwright
Agentic development — Claude Code, Codex, OpenCode, GitHub Copilot, running a multi-agent workflow I designed and use daily
Automation — n8n, Zapier, Make, Power Automate
EU sustainability regulation (CSRD / ESRS, EU Taxonomy) — internal reference point at work for regulatory edge cases, and the reason a lot of my tooling is about turning dense legal text into executable specifications. Before that: comparative European policy research on the Jean Monnet SeBI project.