Video-based facial recognition solution to generate montages of individual fans enjoying the match. Trained various neural network models for classification (2D and 3D), image captioning for natural language descriptions of fans, audio analytics for background noise removal, etc. Organized the process of data collection and labelling, building pipelines for data processing using Luigi. Deploying the models as microservices with Flask. Models are now actively used in production and have significantly reduced the amount of manual labour.