Computer Vision Scientist

About SkillCornerAt SkillCorner, we turn a single broadcast camera feed into rich sports data, 100% automatically. Our Computer Vision and Machine Learning algorithms detect every player, the ball and the referees, locate them on the field, track them frame by frame and recognize who is who. Since 2016, we have grown into a global leader in our field. We cover more than 150 men’s and women’s competitions, hold over 100 billion data points on more than 100,000 professional players, and are trusted by more than 350 football, basketball and American football teams. Clubs, federations and agencies use our data to rethink how they analyze performance, scout and recruit. We are a team of about 150 people around the world, with our tech team based mostly in Paris. The Computer Vision team, 16 people strong, sits at the heart of our mission: it uses state-of-the-art deep learning to deliver precise, high-quality tracking data, at scale and in production. The roleWe are looking for a Computer Vision Scientist to join our Computer Vision team. You will bring solid experience in deep learning applied to computer vision to the models that power our tracking data, from detection and tracking to re-identification and recognition. You will own key models: designing and training architectures, building training and evaluation pipelines, and pushing performance to production-grade quality. You will work with tools like PyTorch/TensorFlow, OpenCV and deep learning architectures (CNNs, Transformers, etc.). You will make technical decisions with real downstream impact and help raise the research and engineering standards of the team. This role sits where applied research meets production. Your work will go from experiment to live deployment, used by professional sports teams and organizations around the world. What you’ll do
• Design, train and optimize deep learning models for detection, tracking, re-identification and recognition
• Own key components of our tracking pipeline, from data and training to evaluation and deployment
• Make and defend technical decisions, grounded in a solid understanding of the theory behind the models you build
• Monitor the state of the art, analyze model performance and turn new ideas into concrete improvements
• Establish the metrics and tooling that measure tracking quality, and drive continuous improvement from one release to the next
• Bring research algorithms into our AWS production pipeline and keep performance stable when new models go live
• Raise the bar on engineering and research practices: code reviews, testing, reproducible experiments and documentation
• Work with Product, Engineering and Operations to turn client needs into concrete tracking improvements
• A graduate degree or PhD in engineering, computer science, mathematics or a related field, with a strong background in mathematics
• Around 3 years of hands-on experience in computer vision with deep learning, in a production environment (exceptional recent graduates are welcome too)
• Real experience in computer vision, not only deep learning in general
• Experience training your own models and a deep understanding of how they work, ideally having modified existing architectures rather than only reusing pre-trained models
• Proven experience with detection, tracking or recognition pipelines, not only research prototypes
• Strong Python for ML: clean, tested, deployable code, with PyTorch or TensorFlow and OpenCV
• The ability to drive technical decisions independently and explain your choices
• Rigor, concise communication, curiosity, autonomy and creativity
• A genuine interest in sport
• Excellent communication skills in English Nice to have
• French: a strong plus, as the team is French-speaking
• Experience with video, object detection and multi-object tracking
• Experience with graph neural networks (GNNs)
• Experience with camera calibration, homography estimation or 3D geometry
• Hands-on experience with AWS workflows and Docker
• Experience running ML pipelines at scale in the cloud (GPU inference, batch processing)
• Publications at top computer vision or ML venues (CVPR, ICCV, ECCV, NeurIPS) Why join us
• Real impact: your work ships to production and is used by more than 350 professional teams worldwide
• Hard, exciting problems: state-of-the-art computer vision on real broadcast video, at scale
• Ownership: own key models and shape how our tracking technology evolves
• A great team: passionate people at the crossroads of AI and sport, with a tech team in the heart of Paris
• #1 Phone interview (20–30 min)
• #2 Technical interview (1 h)
• #3 Technical test (1h)
• #F Culture fit interview (1 h)