R&D teams are not in a position to rigorously compare the performances (e.g accuracy, latency, energy or memory consumptions, etc.) of ML models and hardware on which they are deployed as they face: - Tedious tasks way too time-consuming; - Reproducibility issues when replicating past experiments; - Siloed efforts leading to waste and work duplication. Keeping track of the fast-paced evolution of the State-of-the-art is then a real challenge. CodeReef is developing a web-based collaborative playground equipped to build ML workflows portable on any device and share reproducible performance measurements 100x faster than what was possible before. R&D teams can thus select the most efficient ML solutions to deliver new use cases depending on their objectives and constraints. CodeReef is powered by a proven technology concentrating years of research and already used by corporate and academic AI leaders across the world.