Portfolio
Is tourism raising the rent? Donostia, in open data
Seven data stories about how San Sebastián is changing — housing, barrios, people, climate, and the gap between the tourist city and the lived one. Built from six open sources with a reproducible pipeline; the question that started it turned out to have an uncomfortable answer. Written in Spanish.
Draw a digit, watch a LeNet-5 read it
A LeNet-5 trained on MNIST, running live in the browser. Draw a digit with the mouse and watch every layer's activations light up as the signal propagates to a verdict.
3D reconstruction using Gaussian Splatting and NeRF
Capturing real locations as photorealistic assets for LED-volume virtual production. Benchmarked nerfacto, splatfacto and MCMC against RealityCapture and Postshot, and shipped a desktop app that runs COLMAP pose estimation and reconstructs either locally or on Azure.
Brand recognition in football matches
Measures how much screen time each brand gets in a football broadcast. A YOLO detector finds the boards, a DenseNet201 sorts them into 52 brands — despite motion blur, a 40× class imbalance, and LED ads that change weekly. In production on Azure.
ARoundTheWorld: Collaborative AR for Education
ARoundTheWorld is a multiplatform collaborative AR geography application built on the cleAR architecture. It was evaluated with 44 students across three schools and represents the final paper of my PhD, validating that the architecture can produce applications that integrate seamlessly into existing school curricula.
cleAR: Interoperable Architecture for Multi-User AR
cleAR is a modular, interoperable architecture for building multi-user augmented reality applications in education. Designed from the ground up to bridge the gap between AR's potential and its limited classroom adoption, it was the core contribution of my PhD research.
RAG Document Assistant for Financial Services
A retrieval-augmented generation system built for a large European bank, replacing manual SharePoint search with a conversational assistant that answers questions and cites the exact document and page where each answer was found.
MLOps Platform on GCP for a Spanish Media Group
An MLOps platform on Google Cloud for one of Spain's largest media groups. Took a portfolio of disconnected models — churn prediction, article recommendation and more — and rebuilt them as Vertex AI pipelines with versioning, monitoring and drift detection.
TRACTION: Opera Co-creation for Social Transformation
TRACTION was a Horizon 2020 project, coordinated by Vicomtech, that used opera as a vehicle for social inclusion. My work centred on the Co-creation Stage, the real-time distributed performance tool, spanning development, user-requirements gathering, evaluation, and direct engagement with artistic and community partners across Barcelona, Leiria and Ireland.
Multi-modal Audience Engagement Measurement System
A multi-modal system that measures audience engagement at live events by fusing computer vision with WiFi/Bluetooth signal analysis. Designed, implemented, and validated at Vicomtech in 2020, and released as open-source software.
UFC Fighter Tracking: Multi-Modal Sensing in the Octagon
Real-time analytics for live UFC events: stereo vision in the truss above the octagon, accelerometers in the gloves, GPU inference at the venue. Demoed live during Werner Vogels' keynote at AWS re:Invent 2017.
MPI for Brain Research: Software for Reptilian Neuroscience
Two years as a scientific software developer in Gilles Laurent's department at the Max Planck Institute for Brain Research in Frankfurt, building MATLAB analysis pipelines and R/Shiny web apps for electrophysiologists studying sleep in bearded dragons and visual processing in ex-vivo turtle brains.
inEvent: Structuring and Linking Multimedia Archives of Lectures and Meetings
An EU FP7 project treating archives of lectures and meetings as interconnected hyper-events. At Fraunhofer HHI I led the video analysis: segmentation, slide-transition detection, and face features for speaker linking. Published at VISAPP, Interspeech and the ACM MM Grand Challenge.
AVATecH: Automated Annotation of Audio/Video Corpora for Humanities Research
AVATecH was a joint Fraunhofer / Max Planck project that brought state-of-the-art audio and video pattern recognition into ELAN, the de-facto annotation tool used by linguists, anthropologists, and psychologists worldwide, turning weeks of manual labelling into minutes of supervised review.