Machine Learning Systems
A four-volume textbook and curriculum from Harvard's CS249r course on engineering machine learning systems, paired with hands-on labs, a from-scratch framework, hardware kits and a simulator.
Machine Learning Systems is a Harvard CS249r textbook and curriculum on building efficient, reliable and safe intelligent systems, covering foundations, scaling, agentic AI and physical AI across four volumes, with a print edition planned from MIT Press.
The repository combines the textbook with several companion tools: TinyTorch, a machine learning framework students build themselves from scratch; interactive labs that reason about infrastructure trade-offs using MLSys·im, a modeling engine for infrastructure that is otherwise hard to access; hardware kits for hands-on deployment constraints; and StaffML for checking understanding, alongside an AI-guided reading and quiz layer called Socratiq.
It is written mainly in Python and includes materials for instructors, such as slides and an instructor hub.