How AI workloads actually run on hardware — architecture, CUDA/HIP, libraries, optimization, and the assembly underneath.
A full self-paced C++ track, beginner to advanced: memory and pointers, move semantics, OOP and polymorphism, templates, the STL, data structures and algorithms, design patterns, and tooling.
Python fundamentals, performance/optimization patterns, and PyTorch essentials.
From SELECT to query planning and pgvector — a self-paced SQL track, beginner to advanced, with PostgreSQL as the primary engine.
Not just a chat window — notes on using Claude Code, agents, skills, and MCP as real engineering tools.
A full self-paced ML track, beginner to advanced: regression, decision trees, gradient descent, regularization, ensembles, evaluation, and the classical theory behind why models generalize.
Practical, code-first deep learning with PyTorch: tensors, training loops, CNNs, transfer learning, ResNet/DenseNet, transformers built from scratch, and deployment.
Large language models from first principles: tokenization, embeddings, the Transformer architecture, prompt engineering, RAG, fine-tuning, and GPU-level inference/serving internals.
Time series analysis and quantitative trading from first principles: stationarity, classical forecasting, walk-forward validation, and the leap from a good model to a working trading system.
A self-paced project management track, beginner to advanced: charters, stakeholder analysis, scope and SMART goals, methodologies, planning, and Agile.
A self-paced system design track, beginner to advanced: consistency models, the CAP theorem, scalability building blocks, and worked interview-style case studies.
Day-to-day git commands and a GitHub reference — cloning, branching, remotes, and the workflows that come up constantly.
Editor and terminal-workflow notes, starting with a practical vim reference.
Operating-system fundamentals — tmux for terminal multiplexing, and a beginner-to-advanced Ubuntu terminal command reference.
Language-agnostic debugging and profiling — the skills that matter once the code doesn't work or doesn't run fast enough.