// ai · computer vision · machine learning
Miguel Mendez
Senior ML Engineer at Hudl. I build production computer vision systems that turn football broadcasts into tracking data.
Bio
I am a Senior Machine Learning Engineer at Hudl, based in Vigo, Spain. I build production computer vision systems for sports analytics, with expertise in multi-object tracking, camera calibration, object detection, and generative models. My work spans model research, deployment, and the pipelines that turn match footage into tracking data.
Previously, I developed sports analytics systems at Statsbomb, researched computer vision at Gradiant, and built machine learning solutions for retail at Desigual.
I hold an M.Sc. in Artificial Intelligence from the Polytechnic University of Catalonia, with a year at Purdue University, and a B.Sc. in Computer Science from the University of A Coruña.
News
-
StatsBomb 360 adds AI-powered player ID
Hudl upgraded StatsBomb 360 with AI-powered player identification: every player in each 360 freeze frame is now named and linked to the event data. Built on the broadcast tracking I've been developing at Hudl.
-
Hudl Raw Tracking (XY) Data API ships
Hudl released its Raw Tracking (XY) Data API! Automated player tracking from single-camera broadcast footage at scale (120+ competitions). I've been working on the computer vision behind it since joining Hudl.
-
Statsbomb joins Hudl
Statsbomb have been acquired by Hudl! I am excited to continue working on the future of sports analytics with the Hudl team.
-
Joining Statsbomb as ML Engineer
I am joining Statsbomb as a Machine Learning Engineer.
-
Drone vs Bird paper published in Sensors
Drone vs Bird paper is out! Check our awarded solution.
-
Pyodi goes open source
We have open sourced Pyodi our Python Object Detection Insights library.
-
Winners of the Drone vs Bird Challenge
We have won 2020 Drone vs Bird Detection Challenge.
-
AVSS 2019 Drone-vs-Bird paper out
Paper from IEEE AVSS2019 Drone-vs-Bird Detection Challenge is out.
Posts
-
Managing Multiple GitHub Accounts with SSH
Easily switch between GitHub accounts using different SSH keys
-
LLM Preference Alignment
Understanding PPO, DPO, and ORPO
-
Image Normalization: Comparing CPU vs GPU Performance in PyTorch
Understanding the performance differences in Pytorch's image normalization
-
Httpx vs Aiohttp: Handling High-Concurrency Requests in Async Python
How switching from httpx to aiohttp resolved my httpx read errors and networking issues
-
Understanding Object Tracking Metrics
Comparing MOTA, IDF1, and HOTA for Multi-Object Tracking Evaluation
-
A Guide to Horizontal Pod Autoscaler
Understand and visualize how Kubernetes HPA works with a real world example
- See more posts
Vitæ
-
Supporting automated player tracking across 120+ football competitions through Hudl's Raw Tracking (XY) data. Driving the computer vision work that extracts tracking data from single-camera broadcast footage and aligns it with event data. Designing and deploying multiple deep learning models that work together end-to-end for player tracking, detection, and camera calibration. Contributing across the full stack of the system, from model research to production pipelines processing matches every week.
hudl.com -
Architecting and deploying advanced computer vision systems for automated sports analytics, processing thousands of matches across 90+ leagues. Building custom deep learning models (PyTorch) for real-time player tracking, ball detection, and camera calibration. Established an end-to-end MLOps pipeline for model training, deployment, and monitoring at scale, while leading AI infrastructure modernization and technical initiatives in sports data collection.
statsbomb.com -
Led development of computer vision solutions for real-time object detection and semantic segmentation with deep learning frameworks (PyTorch/TensorFlow). Optimized models for deployment on edge (Nvidia Jetson) and cloud platforms (AWS/Azure). Directed AI best practices, MLOps pipelines, and infrastructure modernization.
gradiant.org -
Designed end-to-end machine learning solutions for retail optimization, including demand forecasting and intelligent inventory management across a global store network. Developed a CNN-based visual search system for product recommendations and style matching. Implemented predictive analytics to enhance stock distribution and improve supply chain efficiency.
inno-it.es -
Researcher at Purdue Datalab, collaborating with PhD candidates on AI projects. Specialized in time series analysis and feature engineering, with thesis: Feature Construction and Classification of Time Series Data. Completed advanced coursework in machine learning and data mining.
purdue.edu -
Specialized in computer vision and neural network architectures, with focus on supervised and unsupervised learning methodologies. Core coursework included deep learning, statistical modeling, and pattern recognition. Developed strong foundation in fundamental AI algorithms and their practical applications.
upc.edu -
Awarded competitive Erasmus and NILS scholarships for international studies. Completed thesis project on computer vision applications, implementing optical flow algorithms for motion analysis using action camera data. Focus on real-time video processing and computer vision techniques.
uis.no -
Comprehensive foundation in computer science fundamentals, with focus on algorithmic complexity, data structures, and software development methodologies.
udc.es