From neural networks to modern language models, explore how LLMs actually work, what they can do, and how to leverage them effectively. This course covers the technical foundations, agentic systems, hardware considerations, and real-world implementation strategies.
Chapters
AI, but not really!
Why we say AI when we mean LLMs, what this course covers, and the terminology trap everyone falls into.
How we got to modern LLMs
The evolution from neural networks to today's language models. Covers the hardware revolution, big data explosion, transformers, attention mechanisms, GPT evolution, training, fine-tuning, and context windows.
Capabilities and constraints
How LLMs process tokens, recognize patterns, and solve problems. Includes multimodal capabilities (vision, audio), best use cases, and fundamental limitations.
Your tireless coding companion
What makes agents different from simple LLM calls. Covers agentic pipeline architecture, tool use, context management, memory systems, and orchestration.
Hardware and infrastructure economics
GPU vs CPU tradeoffs, memory needs, cloud vs self-hosting economics with real calculations, and when each approach makes sense.
Navigating the LLM ecosystem
Open-source models, inference servers, fine-tuning frameworks, orchestration tools, and observability. Understand the stack layers and compose them into working applications.
Tales from the trenches
GitHub Copilot, customer support automation, content generation, and internal tooling. What worked, what didn't, and lessons learned.
The road ahead
Emerging capabilities, agentic system evolution, cost trends, regulatory landscape, and how to prepare for what's next.
Goals
Evaluate when transformers solve problems RNNs cannot
Apply attention mechanism concepts to understand model capabilities
Apply scaling laws to make model size vs training data decisions