Curriculum · AI for Development
Four modules
Build real software with AI. Each module opens on its own course index — start at the top and work down, or jump to the one you need.
How the modules build on each other
AI-Powered Software Development
Build software with AI as an engineering discipline — the agent loop, context engineering, the cost model underneath it, and the verification gates that make any of it safe to ship.
- Reason about a coding agent as a loop, not a chatbot
- Engineer context by volatility, and know what it costs
- Design verification gates an agent cannot talk its way past
Deep Learning Foundations
Neural networks from first principles — regression and gradient descent, dense nets, convolutions for vision, and recurrence for sequences.
- Derive linear and logistic regression from scratch
- Trace forward and backward passes by hand
- Build a CNN and an RNN from scratch
Transformers, LLMs & Agents
The Transformer end to end, a GPT-2 built from tokens up, and a multi-agent research assistant grown from a bare LLM loop.
- Derive self, multi-head, and cross attention
- Build and train a decoder-only GPT-2
- Coordinate planners, critics, and parallel executors
System Design – Distributed Systems
Core distributed-systems concepts — scalability, caching, CAP, message queues — and a case-study-driven interview playbook.
- Reason about scalability, reliability, and availability
- Apply load balancing, caching, CDNs, and CAP
- Structure system design interviews with real case studies