# Project Adam **Domain:** Deep Learning / Distributed Training / Microsoft Research **Doc Type:** Technical Project Node **Maturity:** Developed ## Definition **Project Adam was a Microsoft Research distributed system for training large deep-neural-network models on commodity servers.** The 2014 OSDI paper describes whole-system co-design across computation and communication, extensive use of asynchrony, and large-scale visual-recognition experiments. ## Crown Jewel position [[articles/Pax Silica's Crown Jewel|Pax Silica's Crown Jewel]] places Project Adam in the pre-frontier-model lineage of scalable machine intelligence. Its significance in the article is architectural: model performance depends on training-system design, communication, scheduling, and infrastructure—not only on an abstract set of weights. That documented technical history supports the article's [[wiki/Substrate Determinism|substrate-determinism]] vocabulary without proving the article's broader geopolitical interpretation. ## Routes [[wiki/Microsoft Research|Microsoft Research]] · [[wiki/Project Brainwave|Project Brainwave]] · [[wiki/Distributed Data Architectures|Distributed Data Architectures]] · [[wiki/Deep Learning|Deep Learning]] · [[wiki/Substrate Determinism|Substrate Determinism]] ## Sources / Provenance - Microsoft Research, [Project Adam: Building an Efficient and Scalable Deep Learning Training System](https://www.microsoft.com/en-us/research/publication/project-adam-building-efficient-scalable-deep-learning-training-system/), OSDI 2014. - Microsoft Research, [Introducing Project Adam](https://www.microsoft.com/en-us/research/?p=191050), July 14, 2014.