Andrés Matesanz

Designing and shipping AI-powered automation to radically accelerate internal product workflows — cutting production costs by 90% while preserving quality at scale.

Independent R&D — applied cutting-edge deep learning research to solve a specific client problem, owning the full pipeline from architecture design to evaluation. (NDA — further details confidential.)

Full curriculum design and delivery for Machine Learning bootcamps — taking students from fundamentals to production-ready skills across AI, MLOps, and software engineering.

Launched AI products from zero to demo across multiple industries. Owned the full lifecycle — research, prototyping, model development, and delivery — consistently shipping under tight deadlines and ambiguous requirements.

Built and deployed computer vision systems on embedded hardware for real-world field operations — bridging R&D with production-grade solutions in challenging environments.

Applied research to real-world impact — designed and built AI prototypes that went from academic concept to working field deployment, with results published at an international conference.

Co-founded a startup and shipped the product end-to-end as CTO — from architecture and development to launch. A formative experience in product ownership, speed of execution, and knowing when to pivot.

Reinforcement Learning, Generative Models, Big Data, Evolutive Algorithms. Graduated with Honors — Thesis: 3D MRI Super-Resolution using GANs (4× image quality improvement).

Virtual Reality, Augmented Reality, game theory and procedural generation using Unreal Engine 4.

Physics, mathematics, neurology and biology of vision. Erasmus exchange at Bicocca University (2013–2014). 2nd place at Cooper FORCE Competition 2015.

Live-tracked a real basketball in 3D at one of Spain's most prominent tech festivals using locally-running computer vision models and Rerun for real-time visualisation. All on-device, zero cloud.

Live-built a fully functional generative AI avatar bot (Telegram + Make.com + HuggingFace Spaces) in 30 minutes with zero code written, at the II Congreso de IA de Andalucía.

Custom YOLOv3 model trained from scratch to detect a specific cat — shown live at my first national congress. Covered end-to-end object detection: data collection, training, and mAP validation.