Learning Series Directory
Deep-dive, multi-part guided masterclasses and tutorials taking you from setup and fundamentals to production systems.
Guided Learning Series
Step-by-step sequential tutorials and masterclasses where architectures are built from a blank file to full convergence. Step-by-step sequential tutorials and masterclasses where frontier neural architectures are built from a blank file to full convergence. Step-by-step sequential tutorials and masterclasses where projects and architectures are built from a blank file to full production standards.
Code-First Architecture: Every series is designed to be built from scratch without external black-box libraries, cementing fundamental mathematical operations into production-grade PyTorch. Code-First Architecture: Every series is designed to be built from scratch without external black-box libraries, cementing fundamental principles and engineering practices into production-grade code.
Active & Upcoming Series
Active Implementation Series
Complete, hands-on build series with end-to-end code implementations and training loops. Complete, hands-on build series with end-to-end code implementations and practical workflows.
Cross-Disciplinary Masterclasses
In-depth mathematical derivations paired with interactive visual architectures and developer blueprints.
Diffusion Models Deep Dive
Mathematical derivation of SDEs, score matching, and building a mini conditional latent diffusion pipeline.
PyTorch Developer Blueprint
Production reference covering tensors, autograd computation graphs, custom layers, mixed precision (AMP), and distributed training.
Attention & Sequence Visualizers
Explore interactive multi-head attention weights, positional embeddings, and recurrent flow simulations.