Phase 4: Large Language Models

import { Card, CardGrid } from ‘@astrojs/starlight/components’;

Phase 4: Large Language Models

Learn exactly how ChatGPT, Claude, Gemini, and Llama work — from the moment you type a prompt to when you receive a response.

Difficulty 🟡 Intermediate

Estimated Time 18 Hours

Reading Time 12 Hours

Projects 3

Interview Importance ★★★★★

Prerequisites Deep Learning (Phase 3)

Learning Roadmap

flowchart TD
    M1["📖 Module 1<br/>LLM Foundations"] --> M2["🏗️ Module 2<br/>Transformer Architecture"]
    M2 --> M3["🎓 Module 3<br/>Training"]
    M3 --> M4["⚡ Module 4<br/>Inference"]
    M4 --> M5["✅ Module 5<br/>Revision & Project"]

    M1 --> M1D["What is an LLM? · Tokenization · Context Window"]
    M2 --> M2D["Self-Attention · QKV · Multi-Head · Positional Encoding · GPT Architecture"]
    M3 --> M3D["Pretraining · SFT · RLHF · DPO"]
    M4 --> M4D["Decoding · Temperature · Streaming · Function Calling · Structured Output"]

    style M1 fill:#3b82f6,color:#fff
    style M2 fill:#8b5cf6,color:#fff
    style M3 fill:#f59e0b,color:#fff
    style M4 fill:#22c55e,color:#fff
    style M5 fill:#ef4444,color:#fff

Modules Overview

Learn what LLMs are, how language models work, how text becomes tokens, and why context windows matter.

**4 Lessons · ~3 Hours** Dive deep into the Transformer — self-attention, QKV, multi-head attention, positional encoding, feed-forward networks, decoder-only design, and GPT architecture.

**8 Lessons · ~6 Hours** Understand how LLMs are trained — pretraining, next-token prediction, supervised fine-tuning, RLHF, and DPO alignment.

**5 Lessons · ~4 Hours** Master inference — how LLMs generate text, decoding strategies, temperature, streaming, function calling, structured output, and hallucinations.

**7 Lessons · ~4 Hours** Solidify your knowledge with a cheat sheet, practice questions, MCQs, interview preparation, and a hands-on mini project.

**Revision · ~1 Hour**

Topics Covered

#	Topic	Module	🔥
01	What is an LLM?	M1: Foundations	🔥 Must Know
02	How Language Models Work	M1: Foundations	🔥 Must Know
03	Tokenization	M1: Foundations	🧠 Core Concept
04	Context Window	M1: Foundations	🧠 Core Concept
05	Transformer Overview	M2: Architecture	🔥 Must Know
06	Self-Attention	M2: Architecture	🔥 Must Know
07	Query, Key, Value	M2: Architecture	🧠 Core Concept
08	Multi-Head Attention	M2: Architecture	🧠 Core Concept
09	Positional Encoding	M2: Architecture	🧠 Core Concept
10	Feed-Forward Network	M2: Architecture	🧠 Core Concept
11	Decoder-Only Transformers	M2: Architecture	🧠 Core Concept
12	GPT Architecture	M2: Architecture	🔥 Must Know
13	Pretraining	M3: Training	🔥 Must Know
14	Next Token Prediction	M3: Training	🔥 Must Know
15	Supervised Fine-Tuning	M3: Training	🧠 Core Concept
16	RLHF	M3: Training	💼 Production
17	DPO	M3: Training	💼 Production
18	Inference	M4: Inference	🔥 Must Know
19	Decoding Strategies	M4: Inference	🧠 Core Concept
20	Temperature, Top-K & Top-P	M4: Inference	🧠 Core Concept
21	Streaming	M4: Inference	💼 Production
22	Function Calling	M4: Inference	💼 Production
23	Structured Output	M4: Inference	💼 Production
24	Hallucinations	M4: Inference	🧠 Core Concept
25	Phase Summary	M5: Revision	✅ Revision

What You’ll Build

Project	Module	Description
LLM Token Visualizer	M2	Build a tool that shows how text is split into tokens
Temperature Playground	M4	Interactive tool to compare LLM outputs at different temperatures
Mini ChatGPT	M5	Build a simple CLI chat application using an LLM API

Progress Tracker

0 / 24 Lessons Complete

Getting Started

Ready to master Large Language Models? Start with Module 1 to build your foundation, or jump to any topic that interests you.

Start Learning: