Production AI Engineering — Class Notes

Notes for 26 released classes in the Krish Naik Academy course, covering approximately 93 hours of teaching from 19 July 2026 to 25 October 2026.

Each chapter includes explanations in teaching order, supplied code excerpts, lecture diagrams, live Q&A, key pointers, and a practice checklist. The collection contains 1110 Q&A entries and 103 diagrams.

Use the sidebar or the table below to open a class. Search the notes using the search box, or open the quick revision guide.

Class Topic Recording length
01 Induction and the Research-to-Production Roadmap 5hr 14min
02 Software Setup and the Engineering Workflow 3hr 41min 23s
03 PyTorch Fundamentals — From Data to a Trained Model 4hr 26min 56s
04 Transformers 101, Part 1 — Embeddings, Context, and Self-Attention Intuition 4hr 11min 34s
05 Transformers 101, Part 2 — Queries, Keys, Values, and Contextualized Vectors 3hr 45min 39s
06 Multi-Head Attention and the Transformer Encoder 4hr 1min 50s
07 Decoder Masking, Positional Encoding, and Cross-Attention 4hr 19min 27s
08 Annotated Transformer Implementation — Designing the Encoder and Decoder 3hr 44min 8s
09 Assembling and Training a Transformer from Scratch 3hr 21min 49s
10 Training and Evaluating the Transformer Translation Model 3hr 34min 57s
11 Tokenization Foundations — BPE, Bytes, and Model Compatibility 5hr 17min 55s
12 Byte-Level BPE, WordPiece, and SentencePiece 3hr 22min 51s
13 Custom Domain Tokenization and Vocabulary Trade-offs 3hr 54min 16s
14 Offset Mapping, Custom BPE Decisions, and Prompting Foundations 3hr 22min 54s
15 Prompt Engineering — Reasoning, Structured Outputs, and Evaluation 3hr 25min 9s
16 Structured Generation with Instructor and Outlines 3hr 24min 41s
17 DSPy Part 1 — Signatures, Modules, and Code Execution 3hr 29min 53s
18 DSPy Part 2 — Bootstrap, MIPROv2, and Reflective Prompt Evolution 4hr 11min 16s
19 Fine-Tuning Encoder, Decoder, and Encoder-Decoder Models 2hr 53min 19s
20 KV Cache — Inference, Memory Math, and Multi-Query Attention 3hr 25min 10s
21 Attention Variants — GQA, PagedAttention, and the MLA Introduction 2hr 56min 56s
22 Multi-Head Latent Attention and Rotary Position Embeddings 3hr 8min 57s
23 FlashAttention and PyTorch SDPA — Memory-Efficient Attention Execution 3hr 2min 9s
24 LLM Scaling Laws - Model Size, Data and Compute 1hr 34min 22s
25 AI Application Engineering and RAG Foundations 2hr 48min 40s
26 Building a LangChain RAG HR Assistant 3hr 17min 5s

Downloads and sources

These independent study notes cover the released sessions in this snapshot. Instructor notebooks, diagrams, and technical references are linked in the corresponding chapters. Unresolved questions, failed demonstrations, and deferred topics remain identified. The broader future course roadmap extends beyond these 26 classes.