What you'll learn

Building machine learning models at scale introduces a range of performance and infrastructure challenges. In this course, Distributed Computing for ML, you’ll gain the skills to design, deploy, and optimize scalable machine learning workflows across multi-node environments. First, you’ll learn how to set up a distributed cluster using Ray and PyTorch—from simulating a local cluster to training models across multiple processes. Next, you’ll examine key performance factors such as resource utilization, data partitioning, and communication tradeoffs between processes. Finally, you’ll implement optimization techniques including Distributed Stochastic Gradient Descent (DSGD), experiment with communication backends like Gloo and NCCL, and tune cluster topologies for better performance. You’ll also explore advanced strategies like integrating remote GPU nodes, applying gradient compression, and benchmarking I/O efficiency. When you’re finished with this course, you’ll have the skills and knowledge needed to build and monitor distributed machine learning pipelines on both local and remote infrastructure.