Explore memory optimization techniques for machine learning in this 36-minute conference talk from SNIA Storage Developer Conference 2024. Dive deep into the memory requirements of ML tasks and discover cutting-edge strategies for efficient memory consumption. Understand memory footprints of typical ML data structures and algorithms, including allocation and deallocation during training phases. Master memory-saving techniques like data quantization, model pruning, and efficient mini-batch selection that preserve model performance while reducing resource usage. Learn how to optimize memory across various hardware configurations including CPUs, GPUs, and ML accelerators. Netflix's Tejas Chopra presents practical insights into ML model memory optimization techniques, current memory usage patterns, and emerging research directions in the field.