Applied GenAI: RAG, Vector Databases, Workflow Automation, and Custom Agents

This course is built around one real, running system: the AI chatbot on this very portfolio site. Every technique here — embeddings, vector search, retrieval-augmented generation, workflow automation, and agents — is explained, then shown exactly as it was implemented in production, in Go, against a free-tier LLM. Every section below includes a hands-on element: an interactive visualizer, a runnable-shaped code lab, or a small simulation you can click through. By the end, you won't just understand RAG conceptually — you'll have touched the mechanics, and read the actual shape of a live chatbot built with these exact pieces.

What "Generative AI" actually means for builders

A large language model (LLM) is a next-token predictor trained on enormous text corpora. Given a sequence of tokens, it outputs a probability distribution over the next token, samples from it, and repeats. That's the entire mechanism. Everything else — chat formatting, "reasoning," tool use, agents — is scaffolding built on top of that one operation.

Two consequences matter immediately for anything you build:

• LLMs have no persistent memory of your data. A model trained in 2025 knows nothing about your product, your courses, or your profile written in 2026. It also has no access to anything outside the text you hand it in the prompt.
• LLMs will confidently invent facts (hallucinate) when asked something outside their training data or the prompt's context. There's no internal "I don't know" signal — the model always produces the statistically likely next token, whether or not that token is true.

Retrieval-Augmented Generation exists to solve exactly these two problems: give the model real, current, relevant facts inside the prompt, and instruct it to answer only from those facts.