Aerospace and Mechanical Engineer (UC3M · TU Delft) turned Applied AI Builder
Moving bits to move atoms.
8+ years solving multi-million-dollar operational problems at Amazon and McKinsey. Now independently shipping production AI systems and robotics projects. I bring operational instinct for what AI needs to do for real users, plus hands-on engineering across the full stack.
AI and Agents: End-to-end RAG pipelines, autonomous LLM agents, multi-agent systems, LangChain, ChromaDB, prompt engineering, Langfuse observability
Voice AI: Retell.ai, Voiceflow, ElevenLabs, low-latency multi-turn voice agents with live backend data retrieval
Computer Vision and ML: YOLOv8 fine-tuning (97% mAP@50), ByteTrack, OpenCV, PyTorch, Weights and Biases
Robotics: ROS2, Arduino hardware, PID control, IMU filtering, LIDAR navigation, VLA models
Backend and Cloud: FastAPI, Next.js, Docker, GCP, AWS, Neon Postgres, Vercel serverless
Languages: Python, TypeScript, C++, SQL
| Project | What it is |
|---|---|
| Portfolio Chat Agent | Production RAG agent built with LangChain, ChromaDB, and FastAPI. Knowledge base redesigned around semantic story-level chunking for optimal retrieval. Deployed via Docker to Hugging Face Spaces with conversational memory, rate limiting, and Langfuse observability |
| FastPay AI | Autonomous 3-way invoice matching agent (Next.js, TypeScript, Neon Postgres). Extracts data from messy vendor PDFs with Gemini 2.5 Flash, cross-references POs and WMS receipts, and autonomously approves or flags discrepancies. 7 custom tools, deterministic decision rules, real-time tracing UI, and a full eval suite via Langfuse |
| AI Voice Customer Service Agent | Production voice agent using Retell.ai and Voiceflow, integrated with real-time order lookup via Vercel serverless functions. Low-latency multi-turn conversation with live backend data retrieval |
| Project | What it is |
|---|---|
| Custom YOLOv8 Tracking and Spatial Logic | Custom detection pipeline annotated in Roboflow. Fine-tuned YOLOv8n (~97% mAP@50) with real-time multi-object tracking via ByteTrack and OpenCV. Advanced spatial logic for moving conveyor belts with point-in-polygon pick zones |
| VLA AI Model in 3D Printed Robotic Arm | Physical 4-axis robotic arm integrated with a cloud-based Vision-Language-Action reasoning engine. Uses Gemini 1.5 Pro to translate natural language voice commands into precise mechanical actuations |
AI Engineer or Robotics Engineer roles at ambitious startups, or Technical PM positions where engineering depth is an asset.
Based in Madrid. Open to remote.