In elite swimming, the difference between gold and silver is measured in milliseconds. To keep Australia on top, Swimming Australia and Mantel built "Training Insights" on Databricks - an AI analytics platform that turns the training pool into an F1 garage. This session explores the architecture of a Computer Vision ensemble on Databricks: how a single high-angle camera captures thousands of data points per second - velocity, stroke rate, turn mechanics - processed via an edge-to-cloud pipeline. Key takeaways: Real-time Computer Vision: orchestrating ML models on Databricks to track multiple athletes in a high-glare, splash-heavy environment Unified telemetry: Unity Catalog as a single source of truth across athlete metrics and video metadata Edge-to-cloud: balancing high-speed local processing with scalable cloud model retraining Impact: from manual stopwatches to proactive, data-driven adjustments that provide athletes with real-time feedback and influence Olympic medal outcomes.