Udemy
Turn what you know into an opportunity and reach millions around the world.
Learn More
Your cart is empty.
Keep shopping
Large Language Models: From Foundations to Transformers
Rating: 4.8 out of 5(10 ratings)
156 students

Large Language Models: From Foundations to Transformers

Master the Core Principles Behind Modern AI: deep learning, embedding space, vector search, attention mechanics, and LLM
Created byDr. Ryan Rad
Last updated 7/2026
English
English [Auto],

What you'll learn

  • Master core AI & DL mechanics—from backpropagation and activation functions to self-attention and positional encodings.
  • Implement and benchmark sequence models, contrasting recurrent RNN/LSTM networks with parallelized Transformer architectures.
  • Build dense vector embeddings and deploy scalable semantic vector search engines using industry frameworks like FAISS.
  • Understand the LLM lifecycle: scaling laws, pre-training, Supervised Fine-Tuning (SFT), and preference alignment (RLHF/DPO).
  • Control LLM generation via decoding parameters (Top-K, Top-P, Temp) and inspect open-weights model mechanics to fix errors.

Course content

6 sections • 36 lectures • 7h 5m total length
  • Welcome - Course Introduction7:48
  • Lesson 1.1 - The Evolution of Artificial Intelligence11:20
  • Lesson 1.2 - Rule-Based Systems vs. Learning-Based Systems13:30
  • Lesson 1.3 - Types of Machine Learning6:07
  • Lesson 1.4: Understanding the AI terminology Landscape18:35
  • Lesson 1.5: Foundation of Machine Learning Algorithms23:33
  • Lesson 1.6: Model Evaluation & Validation Fundamentals17:41
  • Introduction to AI and Machine Learning
  • Lab 1 - Machine Learning - From Bits to Beats22:15

    Lab Resources & Structure

    Each lab includes three components designed to guide you from concept to implementation:

    • Deep-Dive Podcast Discussion: An AI-generated podcast overview (powered by NotebookLM) based directly on the course materials, the Agentic AI Book, and curated expert insights, reviewed and refined for technical accuracy. Listen to this first for intuition and context.

    • Hands-On Notebooks: Access the lab via the downloadable .ipynb file to run locally, or open the direct Google Colab link to run it in the cloud without local setup.

    • Practical Tasks: Complete the guided exercises and coding tasks embedded directly inside the notebook to solidify your understanding.

    Instructor Note: All lab code, exercises, and core curricula are intentionally designed and written by the instructor, Ryan Rad. AI is leveraged strictly to generate the companion audio deep dives.

Requirements

  • Intermediate Python: Familiarity with functions, loops, and basic data structures (code is explained step-by-step, but you should be comfortable reading Python).
  • High-School / Basic College Math: Basic familiarity with vectors, matrices, and simple calculus concepts (all key mathematical concepts are refreshed intuitively throughout the course).
  • No prior AI/Deep Learning experience required! We build everything from absolute ground zero.

Description

Agentic AI Mastery: The Complete Learning Path

Master the core architecture driving modern AI—from deep learning and vector embeddings to attention mechanics, Transformers, and LLM scaling.

Your 3-Stage Agentic AI Learning Arc:

  • Course 1: Large Language Models: From Foundations to Transformers

  • Course 2: LLM Engineering: Prompting, RAG, Fine-Tuning, and RLHF

  • Course 3: Agentic AI Engineering: Multimodal, MLOps & Agents

New to the series? Start here! This course provides the essential foundation for the entire learning path.


Why This Course:

Most AI courses jump straight to calling API wrappers, leaving you stranded when models fail or behavior needs fine-tuning. This course takes a different path.

We build your understanding ground-up—from standard machine learning and vector embeddings to the inner workings of self-attention and scaling laws through a carefully structured learning path. While we cover today's essential tools and frameworks, our focus on core principles ensures your knowledge remains valuable as the field evolves.


Course Modules

Module 1: Introduction to AI and Machine Learning

Begin your journey by understanding how machines learn from data. This module demystifies essential AI terminology, contrasts rule-based logic with learning algorithms, and covers fundamental model evaluation strategies—establishing the groundwork needed for deep learning.

Module 2: Deep Learning Fundamentals

Discover why deep learning revolutionized artificial intelligence and learn to think in terms of neural networks. Through hands-on coding and clear intuition, you'll master representation learning, backpropagation, activation functions, and the optimization techniques driving modern neural architectures.

Module 3: NLP Fundamentals and Embedding Space

Explore how computers process and understand human language. This module bridges classical methods like TF-IDF with modern word embeddings, showing you how text is transformed into vector representations to power semantic understanding and real-world vector search applications.

Module 4: Sequential Modeling & Transformers

Trace the evolution of sequence processing from RNNs and LSTMs to the transformer revolution—the architectural breakthrough behind ChatGPT. You'll analyze attention mechanisms, positional encodings, and performance trade-offs while getting hands-on experience building memory and attention models.

Module 5: Large Language Models

Uncover how modern LLMs are built, trained, and fine-tuned. This module covers scaling laws, pre-training, supervised fine-tuning (SFT), preference alignment, and decoding strategies for text generation—setting the stage for advanced LLM engineering and production applications.


This Course Features:

  • 30 Bite-Sized FHD Video Lessons: Clear, structured lectures breaking down complex concepts across 5 comprehensive modules.

  • 5 Hands-on Coding Labs: Implement neural networks, word embeddings, attention mechanisms, and text decoding strategies using industry-standard tools.

  • 5 Deep Dive Audio Podcasts: Reinforce theoretical concepts on the go with custom podcast discussions for every module.

  • Industry-Standard Tools & Ecosystem: Hands-on experience with the production stack used to build, retrieve, fine-tune, and align real LLM systems, including Hugging Face, PEFT, TRL, LlamaIndex, OpenRouter, PyTorch, and Google Colab, working with open-weights models from Qwen, Mistral, Meta (Llama), and OpenAI (GPT-2).

  • Intuitive Visual Diagrams: Complex mechanics—like backpropagation, positional encodings, and self-attention—explained through clean visual architecture diagrams.

  • End-to-End Capstone Project: Build and extend ChefBot—a full-stack AI recipe generator & ingredient vector search app using Streamlit, Hugging Face open-weights models (Qwen 2.5/Gemma 3), SpaCy, and Spotify's Annoy index.

  • 3-Part Series Foundation: The essential baseline to prepare you for advanced engineering in Course 2 (Practical LLM Engineering) and Course 3 (Agentic AI Engineering).


Who This Course Is For:

  • Software Engineers & Developers wanting to transition from standard application development to AI/LLM engineering.

  • Data Scientists & ML Beginners looking to move past traditional scikit-learn pipelines into modern Transformer architectures.

  • Computer Science Students & Tech Professionals who want a deep, principled understanding of how LLMs work under the hood rather than just calling third-party APIs.


Prerequisites:

  • Intermediate Python: Familiarity with functions, loops, and basic data structures (code is explained step-by-step, but you should be comfortable reading Python).

  • High-School / Basic College Math: Basic familiarity with vectors, matrices, and simple calculus concepts (all key mathematical concepts are refreshed intuitively throughout the course).

  • No prior AI/Deep Learning experience required! We build everything from absolute ground zero.


What You'll Master:

Master core AI & DL mechanics—from backpropagation and activation functions to self-attention and positional encodings.

Implement and benchmark sequence models, contrasting recurrent RNN/LSTM networks with parallelized Transformer architectures.

Build dense vector embeddings and deploy scalable semantic vector search engines using industry frameworks like FAISS.

Understand the LLM lifecycle: scaling laws, pre-training, Supervised Fine-Tuning (SFT), and preference alignment (RLHF/DPO).

Control LLM generation via decoding parameters (Top-K, Top-P, Temp) and inspect open-weights model mechanics to fix errors.


Join us on this foundational journey through Generative AI and build the knowledge base that will support your entire career in this rapidly evolving field.


Continue Your Learning Journey

Upon completing this course, you'll be perfectly prepared for:

  • Course 2: Practical LLM Engineering: Prompting, RAG, Fine-Tuning, and Alignment

  • Course 3: Agentic AI Engineering: Multimodal, MLOps & Agents


Module 1: Introduction to AI and Machine Learning

Video Lessons:

  • Lesson 1.1: The Evolution of Artificial Intelligence

  • Lesson 1.2 - Rule-Based Systems vs. Learning-Based Systems

  • Lesson 1.3: Types of Machine Learning

  • Lesson 1.4: Understanding the AI Terminology Landscape

  • Lesson 1.5: Foundation of Machine Learning Algorithms

  • Lesson 1.6: Model Evaluation & Validation Fundamentals

Resources:

  • Module 1 summary notes

  • Quiz 1

  • Lab 1 - Build a Spotify Hit Prediction Engine (Regression & Classification)

  • Lab 1 (Deep Dive Podcast)



Module 2: Deep Learning Fundamentals

Video Lessons:

  • Lesson 2.1: Why Deep Learning? DL vs ML

  • Lesson 2.2: Introduction to Neural Networks

  • Lesson 2.3: Activation Function & Representation Learning

  • Lesson 2.4: Universal Approximation Theorem Explained

  • Lesson 2.5: Backpropagation and Common Challenges

  • Lesson 2.6: Deep Learning Optimization Techniques

Resources:

  • Module 2 summary notes

  • Quiz 2

  • Lab 2 - Build a Deepfake vs. Real Face Classifier on Digital Trust Data (CNN & Transfer Learning)

  • Lab 2 (Deep Dive Podcast)



Module 3: NLP Fundamentals and Embedding Space

Video Lessons:

  • Lesson 3.1: Introduction to Natural Language Processing

  • Lesson 3.2: Text Preprocessing Fundamentals

  • Lesson 3.3: Classical Methods: Bag-of-Words, N-Gram, TF-IDF

  • Lesson 3.4: The Magic of Word Embeddings

  • Lesson 3.5: Advanced Embeddings: Word2Vec to Contextual

  • Lesson 3.6: Vector Search and Real-World Applications

Resources:

  • Module 3 summary notes

  • Quiz 3

  • Lab 3 - Build a Semantic Book Search Engine (TF-IDF, Sentence-Transformers, and FAISS Vector Search)

  • Lab 3 (Deep Dive Podcast)


Module 4: Sequential Modeling & Transformers

Video Lessons:

  • Lesson 4.1: Introduction to Sequential Modeling

  • Lesson 4.2: From RNN to LSTM

  • Lesson 4.3: Transformer - A New Paradigm

  • Lesson 4.4: Positional Encoding

  • Lesson 4.5: Attention Mechanism

  • Lesson 4.6: LSTM vs Transformer - Performance Showdown

Resources:

  • Module 4 summary notes

  • Quiz 4

  • Lab 4 - Build a Creative Book Title Generator (LSTM vs. Transformer Attention Mechanics)

  • Lab 4 (Deep Dive Podcast)


Module 5: Large Language Models

Video Lessons:

  • Lesson 5.1: Language Models - From N-grams to Neural

  • Lesson 5.2: Scaling Laws and the Discovery of LLMs

  • Lesson 5.3: How LLMs Learn: Pre-training, SFT, and Preference Alignment

  • Lesson 5.4: Decoding Strategies for Text Generation

  • Lesson 5.5: Current Limitations and Challenges

  • Lesson 5.6: Preview of LLM Engineering Techniques

Resources:

  • Module 5 summary notes

  • Quiz 5

  • Lab 5 - Build an LLM Decoding & Hallucination Inspection Engine (Qwen 2.5-0.5B, Logits, Top-K/Top-P)

  • Lab 5 (Deep Dive Podcast)

Who this course is for:

  • Software Engineers & Developers wanting to transition from standard application development to AI/LLM engineering.
  • Data Scientists & ML Beginners looking to move past traditional scikit-learn pipelines into modern Transformer architectures.
  • Computer Science Students & Tech Professionals who want a deep, principled understanding of how LLMs work under the hood rather than just calling third-party APIs.