Deep dives into ML research, mathematical intuition, systems design, and software engineering — written for practitioners who enjoy the details.
How a 1959 algorithm for finding shortest paths became a system that answers a billion routing queries a minute.
A self-contained learning project illustrating core Pandas and PySpark concepts through minimal, working code examples backed by synthetic data.
A comprehensive project illustrating deep learning quantization concepts through minimal working code examples, visualizations, and mermaid diagrams.
Ten reinforcement learning algorithm implementations built for educational purposes — from tabular Q-learning through PPO, SAC, and GRPO.
Every major Ray subsystem through minimal, runnable Python examples — the patterns used in production distributed ML systems.
A self-contained, runnable illustration of Apache Kafka concepts and event-driven design patterns, using an in-memory mock broker so no external services are required.
A comprehensive reference for engineering production AI/ML systems — ML system architecture, MLOps, LLMOps, data engineering, and scaling from prototype to production.
ONNX features and graph optimization techniques with minimal, runnable code examples — every concept paired with working code and mermaid diagrams.
25 standalone CUDA C++ tutorials covering GPU programming from fundamentals to advanced library usage, targeting NVIDIA H200 (sm_90) with ablation experiments.
A comprehensive reference spanning architectural paradigms, design patterns, distributed systems theory, reliability, security, and organizational practices at scale.
A from-scratch, explainable implementation of a small vision-language model for image captioning, built entirely with PyTorch.
Core probability and statistics concepts through minimal, runnable Python code and SQL analytics — read the source, understand the math.
A curated collection of minimal, runnable examples illustrating production-grade Python engineering patterns, each reading its configuration from YAML and writing structured logs.
Minimal, hands-on tutorials for understanding Transformer model training and inference on GPU.
Re-deriving scaled dot-product attention from first principles, with annotated PyTorch code, mermaid architecture diagrams, and a discussion of why positional encoding works the way it does.