This lesson assumes no prior knowledge. It’s structured around a single, practical project that integrates fundamental AI concepts (tokens, embeddings, context windows, prompt design) with retrieval and orchestration (RAG, vector databases, LangChain/LangGraph, MCPs, and agents).
| Topic | What it is | Why it matters |
|---|---|---|
| Tokens, context windows, prompt design | The basic units and limits for language model input and strategies for guiding behavior | Impacts cost, capability, and response quality |
| Embeddings | Numerical vectors that represent text semantics | Enables semantic search and similarity-based retrieval |
| Retrieval-Augmented Generation (RAG) | Combining retrieval from a knowledge store with generation by a model | Improves factual accuracy and relevance for LLM outputs |
| Vector databases | Storage and indexing systems for embeddings | Fast, scalable similarity search for RAG pipelines |
| LangChain / LangGraph | Orchestration libraries for composing models, prompts, and tools | Simplifies building complex, multi-step AI workflows (agents) |
| MCPs (Model Context Protocols) | Conventions for how models share context and tools | Helps agents coordinate model calls and external tools |
| AI Agents | Systems that use models + tools to perform tasks autonomously | Enables multi-step, tool-enabled workflows like data lookups, API calls, and reasoning |
- Core AI fundamentals (tokens, embeddings, context windows, prompt design)
- Retrieval-Augmented Generation and vector databases — how embeddings are stored and searched
- Orchestration with LangChain and LangGraph, and how they help build agents
- MCPs and agent coordination across models and tools
- A single end-to-end project that ties these components together
