AI System Design
Eight decisions for designing AI-powered products. User-first design, model selection, orchestration, failure modes.
Agent Teams
Multi-agent orchestration patterns. Agent loops, DAGs, fan-out/fan-in, and the six knobs per agent.
Training Loops
How models learn. Forward pass, backward pass, gradients, optimization, and learning rate schedules.
Model Architecture
Inside the neural network. Embeddings, attention layers, multi-head attention, MLPs, depth vs width.
Data Pipelines
How text becomes training data. Tokenization, batching, packing, and evaluation metrics.