Course Overview

Based on Andrej Karpathy's "Deep Dive into LLMs like ChatGPT" (3h 31m)

The Three-Stage Pipeline

Modern LLMs are built through three distinct stages:

  1. Pre-Training: Learn language patterns from internet-scale data
  2. Post-Training (SFT): Learn helpful assistant behavior from conversations
  3. Reinforcement Learning: Discover novel reasoning through trial and error

Start Learning

01

Pre-Training Fundamentals

45 min
→ 02

Inference & Base Models

30 min
→ 03

Post-Training & Fine-Tuning

45 min
→ 04

Hallucinations & Mitigations

30 min
→ 05

Reinforcement Learning

45 min
→ 06

Future Directions

15 min
→ 07

AI Analytics Platforms

Bonus
→ 08

Memory Systems for LLMs

Bonus
→