Engineer in Madrid. I work on computer vision, multi-agent reinforcement learning and real-time robotics, mostly on edge hardware.
Background: Mathematical Engineering and Artificial Intelligence (iMAT), Universidad Pontificia Comillas ICAI, 2022-2026. My bachelor's thesis built a benchmark for coordinating teams of autonomous aerial platforms against wildfires. A policy trained with teams of 3 to 10 drones also solves it with 15, without retraining. The code release is in preparation.
- TorchRL: added the
ActionScaling(#3765) andFlattenAction(#3766) transforms, closing a feature request open since 2023.
- turtlebot3-ros2: ROS 2 localization (EKF in simulation, particle filter on the real robot) and navigation for a TurtleBot 3.
- AirTrackPad: mouse control with hand gestures from a webcam (MediaPipe, OpenCV and a small gesture classifier).
- Tensor-Networks: CNN compression with Tucker and CP decompositions, plus MPS simulation of quantum circuits.
- A-Lift: reinforcement learning for a warehouse pick-and-place task (SARSA(λ) with tile coding, Q-learning, DQN).
- computer-vision-ii: course projects on transfer learning, YOLO detection, explainability, adversarial attacks and generative models.
- nlp-course-labs: NLP labs: bigram and naive Bayes models, skip-gram word embeddings, transformer encoder and decoder, fine-tuning.
- discrete-math-projects: Discrete Mathematics projects with Jorge Vančo: a number theory toolkit, an RSA chat and a Madrid route planner on a graph library written from scratch.