Your learning path
Understand the foundations. Build working systems. Make your work worth showing.
Build agents you can explain.
Build reliable agent systems through Python labs, failure analysis, and projects backed by evidence.
Python, mathematical intuition, transformers, and your first bounded agent.
Python at the system boundary
Build predictable async programs with explicit data contracts and reproducible handoffs.
The mathematics of useful predictions
Connect vector geometry, probability, optimization, and evaluation to decisions an agent makes.
Neural computation from first principles
Trace a network forward, differentiate it backward, and build attention from weighted sums.
Inside a modern language model
Explain tokenization, transformer data flow, positional rotation, cache memory, and scaling tradeoffs.
Integrating models without losing control
Use the Responses API, structured output, streaming, and explicit function-call orchestration.
Your first bounded agent service
Build a small controller with explicit state, evidence-based completion, and a minimal UI/API boundary.