The Center for Artificial Intelligence Innovation at the National Center for Supercomputing Applications (NCSA) has released a new educational resource aimed at demystifying deep learning frameworks: a hands-on tutorial series on MyTorch, presented by AI researcher and developer Priyam Mazumdar.
MyTorch is a lightweight, educational deep learning library designed to help learners understand the inner workings of neural networks by building them from scratch. The tutorial series walks viewers through the architecture, design principles, and implementation of MyTorch, offering a transparent look into how core components like autograd, layers, and optimizers are constructed.
The tutorial is available in two parts on YouTube:
The accompanying codebase is open-source and hosted on GitHub: https://github.com/priyammaz/MyTorch
Mazumdar, who developed MyTorch as a pedagogical tool, emphasizes the importance of understanding the fundamentals behind high-level frameworks like PyTorch and TensorFlow. “MyTorch is not meant to replace these tools,” he explains in the video, “but to help learners grasp what’s happening under the hood.”
The tutorial is part of the Center for AI’s broader mission to make cutting-edge machine learning education accessible to students, researchers, and practitioners across disciplines.