Generative AI & LLM Engineering: Transformers to Agents

A practical 40-lesson follow-on to "Machine Learning from Zero to Advanced". Go from how GPT models are trained and sampled, through prompting, embeddings, vector search, RAG and fine-tuning, to tool-using agents, MCP and production serving. It ends with four capstone projects and a roadmap. Every lesson gives intuition, math, a worked numeric example and runnable provider-neutral Python.

Lessons: 40
Hours of video: 8.9
Minutes of reading: 417
Finished here: 0

Start lesson 1 Open on YouTube

#	Lesson	Video	Read
1	How GPT Is Trained: Next-Token Prediction Explained with Math and Code	13:58	11 min
2	Scaling Laws Explained: Why Bigger Models Keep Getting Better	12:34	10 min
3	Sampling Explained: Greedy, Temperature, Top-k, Top-p (With Numbers)	13:03	10 min
4	Context Windows and RoPE Explained: How LLMs Know Word Order	13:58	11 min
5	The KV Cache Explained: Why LLM Decoding Is Memory-Bound	13:33	11 min
6	Why LLMs Hallucinate (and What Actually Helps)	11:42	9 min
7	Prompt Patterns and System Prompts: Programming in Plain English	12:49	10 min
8	Few-Shot Prompting and In-Context Learning Explained (With Code)	12:19	10 min
9	Chain-of-Thought and Reasoning Models: How LLMs Learn to Think	12:48	10 min
10	Structured Output and Evaluating Prompts Like Real Software	15:14	12 min
11	Sentence Embeddings and Contrastive Training: How Meaning Becomes Math	12:47	10 min
12	Cosine vs Dot Product: How Exact Vector Search Really Works	13:03	10 min
13	ANN Indexes Explained: HNSW, IVF and Product Quantization	15:29	12 min
14	Vector Databases in Practice: Filtering, Sizing, and Real Code	13:14	10 min
15	RAG Explained: How to Ground LLMs in Your Own Data	12:27	10 min
16	Chunking Strategies for RAG: How to Split Documents Right	13:41	11 min
17	Retrieval and Query Rewriting: Fix RAG Before It Fails	14:39	12 min
18	Reranking with Cross-Encoders: The Cheap Trick That Fixes RAG Search	13:46	11 min
19	Hybrid Search Explained: BM25 + Dense Vectors with Reciprocal Rank Fusion	13:10	10 min
20	Evaluating RAG: Metrics, Test Sets and 8 Failure Modes	12:26	10 min
21	Supervised Fine-Tuning Explained: Chat Templates, Loss Masking and Code	11:47	9 min
22	LoRA Math Explained: Low-Rank Adaptation Derived From Scratch	14:50	11 min
23	QLoRA Explained: Fine-Tune a 7B Model on One GPU with 4-Bit Quantization	12:55	10 min
24	RLHF Explained: Reward Models, KL Penalty and PPO Step by Step	15:02	12 min
25	DPO Derived: Preference Optimization Without RL	14:19	11 min
26	Fine-Tune vs RAG vs Prompt: How to Decide (A Practical Framework)	12:51	9 min
27	Tool and Function Calling: How LLMs Actually Use Tools	13:11	10 min
28	The ReAct Loop, Planning and Reflection: How AI Agents Actually Think	13:18	11 min
29	Agent Memory Explained: Short-Term, Long-Term and Context Management	13:09	10 min
30	Multi-Agent Systems: When They Help and When They Hurt	12:55	10 min
31	Model Context Protocol (MCP) Explained: One Standard for AI Tools	13:24	11 min
32	Latency and Cost Math for LLM Applications: Predict Your Bill	13:51	10 min
33	Quantization Explained: INT8, INT4, GPTQ and AWQ for LLMs	15:13	12 min
34	Serving LLMs with vLLM: Continuous Batching and PagedAttention Explained	14:37	11 min
35	LLM Guardrails and Prompt Injection: Securing and Monitoring AI in Production	14:04	11 min
36	Build a Document Q&A RAG Bot: Hybrid Search, Reranking, Citations, Evals	14:14	11 min
37	Capstone 2: Build a Repo-Aware Coding Assistant (Python, Sandboxed, Tested)	13:05	10 min
38	Capstone: Fine-Tune a Small Open Model End to End (QLoRA + TRL)	12:32	10 min
39	Capstone 4: Deploy a Tool-Using AI Agent to Production (FastAPI, Docker, Guardrails)	12:50	10 min
40	LLM Engineering Roadmap: What You Learned and Where to Go Next	11:04	8 min