Generative AI Architecture
How generative AI systems are built and run: foundation models, prompting, embeddings, retrieval-augmented generation, vector databases, fine-tuning, evaluation, and the GPU/TPU infrastructure underneath. Pairs with Understanding Transformers for the model mechanics and Agentic AI for agents.
Lessons
1
Foundation Models and How They're Served
2
Prompt Engineering as Application Logic
3
Embeddings: Meaning as Geometry
4
Retrieval-Augmented Generation
5
Vector Databases: When You Need One
6
Fine-Tuning vs. RAG vs. Prompting
7
Evaluating AI Systems
8
GPUs, TPUs, and AI Infrastructure