Compilers for Machine Learning (Spring 2027) (10-411.com)

Carnegie Mellon University has announced a new hands-on course, "Compilers for Machine Learning," set to launch in Spring 2027, instructed by notable developer George Hotz. This course will guide students through the process of building a machine learning compiler from scratch, covering crucial topics such as intermediate representation (IR) design, code generation, kernel optimization, and GPU programming. Students will engage in weekly coding assignments that progressively enhance their understanding of compiler technology and machine learning frameworks, culminating in a final project to apply their knowledge practically. This course is significant for the AI/ML community as it addresses the growing need for efficient machine learning compilers capable of optimizing models for diverse hardware architectures, particularly GPUs. With the increasing complexity of AI models, including large language models (LLMs), the ability to translate high-level algorithms into efficient, runnable code is paramount. The course’s focus on practical skills—ranging from memory management to parallelism and automatic differentiation—equips the next generation of engineers with the essential tools to push the boundaries of AI model training and deployment. Participants will emerge with a hands-on understanding that could lead to innovations in AI performance and efficiency.