MODERN AI
Generative AI
Understand how AI learns patterns from data and uses them to create text, images, audio, video, code and multimodal experiences.
Generative AI Learning Path
Start with what generative AI means, then learn how major model families create different kinds of content and how these systems are trained, adapted and used in practice.
What Is Generative AI?
The central question of this entire course: how AI moved from predicting/classifying things to creating new content — building directly on your Machine Learning and LLM foundations, not starting over.
What Does "Generative" Actually Mean?
Going beneath the surface: what a model actually learns from data, what it means to learn a distribution, and why generated content can be really novel while still being built entirely from learned patterns.
Generative AI vs. ML vs. DL vs. LLM
A clear, honest picture of how AI, Machine Learning, Deep Learning, Generative Models, Foundation Models, and LLMs actually relate — deliberately avoiding an oversimplified, perfectly nested hierarchy.
Why Generative AI Became Possible
The real historical and technical story: how larger datasets, better architectures, GPUs, parallel computation, and scaling converged — Generative AI did not suddenly appear, closing out Level 1.
Generative Modeling From First Principles
The foundational distinction between discriminative and generative modeling that underlies every model family in this course — starting Level 2, from data through learned patterns to sampled new outputs.
Autoregressive Generation
The generative model family you already understand best from your LLM course, made fully explicit — token-by-token generation, teacher forcing, training vs. inference, and its extension beyond text.
Variational Autoencoders (VAEs)
A really different generative strategy from autoregressive models: compressing data into a latent representation and learning to reconstruct and generate from it — encoder, decoder, and the intuition behind KL divergence.
Generative Adversarial Networks (GANs)
A completely different training strategy: a generator and a discriminator locked in competition — the counterfeiter vs. detective story, adversarial training, mode collapse, and why diffusion models later became dominant.
Diffusion Models — Core Intuition
The approach that has become dominant for modern image generation: gradually adding noise to data, then learning to reverse that process one small step at a time — deep intuition before any equations, closing Level 2.
Sampling
What it actually means to sample from a probability distribution, and how temperature, top-k, and top-p shape the creativity, diversity, and reliability of generated output — starting Level 3.
Latent Space
The concept first encountered in the VAE module, covered fully in its own right — what latent representations actually are, why they're useful, and how they reappear throughout generative AI, including embeddings and diffusion.
Diffusion Model Architecture
The specific architectural components that make Module 9's core diffusion intuition practical at scale — U-Net, time embeddings, conditioning, cross-attention, and latent diffusion.
Text-to-Image Generation Pipeline
Assembling everything from diffusion intuition through architecture into the complete, end-to-end pipeline from a text prompt to a final generated image — closing Level 3.
Text Generation
Connecting directly back to your LLM course as Level 4 begins — the complete text generation pipeline from prompt to output, and decoding strategy choices in practice.
Image Generation
Moving beyond text-to-image into the broader family of image-generation applications: image-to-image, inpainting, outpainting, editing, and conditioning — what each does and how it works conceptually.
Audio Generation
Extending the generative AI mental model to a new modality: speech synthesis, music generation, and voice cloning — the high-level architecture, without deep specialized audio mathematics.
Video Generation
Confronting the really harder challenges of generating coherent video: temporal consistency, motion, long sequences, and why video generation is fundamentally more difficult than image generation.
Code Generation
A really different kind of generative task: code has verifiable correctness, unlike free-form text, image, or video — completion, transformation, debugging, testing, and why generated code should never be blindly trusted.
Multimodal Generative AI
How modern models reason across multiple modalities simultaneously — shared latent space, cross-modal generation, and combining everything covered across text, image, audio, video, and code — closing Level 4.
Foundation Models
What makes a model a 'foundation model,' the pretrain-then-adapt paradigm, and why this shift changed how Generative AI systems are actually built and deployed — starting Level 5.
Pretraining and Fine-Tuning
Going deeper into the mechanics of pretraining, and a direct, practical decision framework for choosing between prompting, fine-tuning, and RAG for a given real-world task.
Alignment
How a raw, capable pretrained model becomes a really usable, helpful, safe assistant through instruction tuning and alignment techniques — closing Level 5.
GenAI Application Architecture
How real, production Generative AI applications are actually structured and built on top of everything covered so far — starting Level 6: Building Applications.
GenAI Application Stack
A practical look at the categories of tools and technologies that fill out a GenAI application's architectural layers: orchestration frameworks, vector databases, and model providers.
Inference and Model Serving
Going deeper into the model layer: what happens when a request actually reaches a model, and the real infrastructure considerations behind serving generative models at scale — latency, throughput, and batching.
API-Based GenAI, Open-Source vs. Proprietary
The practical decision between using a hosted API and self-hosting an open-source model — cost, control, and operational trade-offs, building directly on inference infrastructure concepts.
Token Economics and Streaming
A deeper look at exactly how usage-based cost is calculated for generative AI, and practical, concrete strategies for managing and optimizing it — closing the cost picture from Module 25.
GenAI + RAG
Revisiting retrieval-augmented generation from the Prompt Engineering course, now framed fully within this course's generative modeling and latent space concepts — why RAG really works.
GenAI + Agents
Revisiting agentic AI from the Prompt Engineering course, now framed through this course's complete generative modeling picture: autoregressive generation, sampling, tool use, and layered application architecture.
GenAI + Prompt Engineering & Context Engineering
Closing Level 6 by directly connecting the entire Prompt Engineering course to this course's generative modeling framework — why prompting works, mechanically, and how context engineering scales it.
GenAI Evaluation
How to systematically measure whether a GenAI application is actually working well, building directly on evaluation principles from the Prompt Engineering course — starting Level 7: Production.
Hallucination in GenAI
A direct, focused treatment of the risk referenced throughout this course: why hallucination happens mechanistically, how to detect it, and genuine, practical mitigation strategies across every modality.
Safety, Responsible GenAI & Guardrails
Broadening from hallucination to the full picture of responsible GenAI deployment — the ethical concerns flagged throughout this course, and practical guardrail design.
Common GenAI Mistakes
A direct, practical catalog of the mistakes flagged individually throughout this entire course, gathered into one comprehensive, actionable reference for building GenAI applications.
When to Use GenAI (vs. Traditional Software/ML)
A really honest look at when Generative AI is NOT the right tool — matching problems to the right solution, whether that's GenAI, traditional software, or discriminative ML.
Cost, Latency, Reliability & Model Selection
Closing Level 7 by tying together every production consideration into one unified framework for choosing the right model and configuration for a given application.
GenAI Design Patterns
Recurring, reusable architectural patterns that combine everything covered across this entire course into proven, practical solutions — starting Level 8: Advanced.
Real-World Case Studies & Practical Projects
Applying everything covered across this entire course to genuine, realistic end-to-end project scenarios — three complete case studies walking through the full design process.
Comparisons and Misconceptions
Directly addressing the field's most persistent points of confusion, gathered and clarified using everything covered across this entire course.
Interview Masterclass & Final Learning Journey
The final module of this course: a comprehensive interview preparation resource covering every level, and a complete recap of the entire 40-module learning journey.
WHY GENERATIVE AI?
Understand how models create new content
Generative AI brings together machine learning, deep learning, language models, diffusion and multimodal systems. Understanding the model families and the trade-offs behind them helps you choose the right approach for real AI applications.