Large Language Models and AI Engineering

You need to sign in or sign up before continuing.

reading · we frame, you read MIT or the canonical taught · we author, no canonical fits ↺ spirals back to earlier lessons

Course locked

Complete Natural Language Processing first.

This course unlocks once you've finished its prerequisite. Open prerequisite →

Prerequisites

Sequence Models and Generative Models not yet

Natural Language Processing not yet

Complete prerequisites first to enroll.

reading · 40 min + ~183 min canonical

Scaling Laws and Pretraining: Why Bigger Became Better

You've arrived at the final course, and it answers the question the whole Deep Learning and NLP arc has been building toward: how did "predict the next word" — the humble n-gram t…

primary Full Stack LLM Bootcamp

preview →

reading · 35 min + ~183 min canonical L1

From Predictor to Assistant: Instruction Tuning and RLHF

The base model from the last lesson — a giant next-token predictor trained on the internet — is not the ChatGPT or Claude you talk to. If you hand a raw pretrained model the promp…

primary Full Stack LLM Bootcamp

after L1 locked

reading · 35 min + ~183 min canonical L1

Prompt Engineering: Programming the Model in English

Here is the strangest and most consequential fact about aligned LLMs: you program them in plain English. There's no API of functions to call, no parameters to tune, no code to wri…

primary Full Stack LLM Bootcamp

after L2 locked

reading · 40 min + ~183 min canonical L1

Retrieval-Augmented Generation and Fine-Tuning: Giving the Model Your Knowledge

Prompting (last lesson) is powerful but hits a hard wall: the model only knows what was in its training data, frozen at its pretraining cutoff. It doesn't know your company's inte…

primary Full Stack LLM Bootcamp

after L3 locked

reading · 35 min + ~183 min canonical L1

Making It Run: Tokenization, Inference, and Serving

You've learned to build with LLMs — prompt them, ground them with retrieval, fine-tune them. Now the unglamorous, load-bearing reality: running LLMs is expensive, slow, and hard t…

primary Full Stack LLM Bootcamp

after L4 locked

reading · 40 min + ~183 min canonical L1

Evaluation, Safety, and Building AI Applications

This is the final lesson — of this course, of Quarter 16, and of the entire four-year computer-science journey — and it's the one that turns everything you've learned into respons…

primary Full Stack LLM Bootcamp

after L5 locked

exercise reading · 15 min L1

Exercises: Easy (Warm-up)

- [ ] Prompt engineering exercises — Write zero-shot, few-shot, chain-of-thought prompts for 5 tasks (classification, extraction, summarization, code generation, reasoning) - [ ] …

after L6 locked

exercise reading · 30 min L1

Exercises: Medium (Core)

- [ ] Fine-tune a model with LoRA — Use PEFT library to fine-tune Llama-2 7B on a custom dataset (< 1000 examples) - [ ] Implement a RAG system with evaluation — Chunk strategy co…

after L7 locked

exercise reading · 45 min L1

Exercises: Hard (Stretch)

- [ ] Quantize a model — Apply GPTQ/AWQ quantization to a 7B model, benchmark speed vs quality tradeoff - [ ] Implement RLHF concepts — Build a reward model from human preferences…

after L8 locked

L10

exercise reading · 45 min L1

Exercises: Hard (Stretch)

- [ ] Red-team an LLM — Systematically find prompt injection, jailbreak, and data extraction vulnerabilities; document and propose mitigations

after L9 locked

L11

exercise reading · 60 min L1

Mini-Project

- [ ] Build a full AI application — RAG-powered chatbot for a specific domain (e.g., Ruby on Rails documentation): document ingestion, chunking, embedding, vector store, retrieval…

after L10 locked

L12

project reading · 360 min L1

Project: RAG Pipeline

Build a retrieval-augmented generation system: chunk documents (experiment with fixed-size vs semantic chunking), embed with sentence-transformers, store in ChromaDB or FAISS, ret…

after L11 locked

L13

project reading · 360 min L1

Project: AI Agent with Tool Use

Build an LLM agent that can: search the web (via API), run Ruby code (sandboxed), query a SQLite database, and read files. Implement a simple planning loop: the agent reasons abou…

after L12 locked

L14

project reading · 300 min L1

Project: LLM Evaluation Framework

Build a Python tool that evaluates LLM outputs on a test set: compute ROUGE, BERTScore, and custom rubric-based scores (using an LLM-as-judge). Support side-by-side comparison of …

after L13 locked

L15

assessment reading · 63 min L1

Self-Assessment: Large Language Models and AI Engineering

- [ ] What is RAG (Retrieval-Augmented Generation)? Explain the pipeline: chunk, embed, store, retrieve, generate. - [ ] What is LoRA? How does it enable efficient fine-tuning? - …

after L14 locked

L16

assessment reading · 60 min L1

Quarter 16 Gate Check: NLP, LLMs, and Applied AI

- [ ] Course 16.1 knowledge check passed - [ ] Course 16.2 knowledge check passed - [ ] AI Code Review Assistant capstone completed

after L15 locked

L17

assessment reading · 120 min L1

Year 4 Assessment: ML and AI

Complete all of the following:

after L16 locked

17 lessons. Read in order; spiral back when you need to. By the end you'll have used the core ideas twice — once on the abstract, once on something you'll meet at work next week.