Learn Gen AI by building it.
Hands-on, self-paced labs covering the Gen AI stack, from deep-learning foundations to production RAG systems. Every lab is about two hours of reading, interactive visuals and real Python you write yourself.
Open modules · start here
+7 more, access restricted:Fine-TuningAI AgentsMultimodal AIApp DevelopmentMLOpsResponsible AICapstoneSee all modules ↓Open to everyone registered for a phaiAI program or hackathon. Sign in with the email you registered with.
Eleven modules, foundations to capstone.
- 014 labs
Foundations of AI & Deep Learning
Build essential skills: Python for AI, neural networks from scratch, deep learning frameworks and training-loop mastery.
- 1.1 Python for AI
- 1.2 Neural Network Fundamentals
- 1.3 Deep Learning with PyTorch/TensorFlow
- 1.4 Loss Functions, Optimizers & Training Loops
- 025 labs
Core Generative AI
Explore generative models: autoencoders, VAEs, GANs and diffusion models for image generation and latent-space exploration.
- 2.1 Introduction to Generative Models
- 2.2 Autoencoders & VAE
- 2.3 Generative Adversarial Networks (GANs)
- 2.4 GAN Variants (DCGAN, WGAN, StyleGAN)
- 2.5 Diffusion Models (DDPM, Stable Diffusion)
- 036 labs
NLP & Large Language Models
Master text representation, transformer architecture, pre-trained models, prompt engineering and building with LLM APIs.
- 3.1 Text Representation (Embeddings, Word2Vec, GloVe)
- 3.2 Transformer Architecture
- 3.3 Pre-trained Language Models (BERT, GPT)
- 3.4 LLM Architecture & Scaling Laws
- 3.5 Prompt Engineering & In-Context Learning
- 3.6 LLM APIs (OpenAI, Anthropic, Open-source)
- 044 labs
Retrieval-Augmented Generation
Build production RAG systems: vector databases, document chunking, pipeline design, hybrid search and RAGAS evaluation.
- 4.1 Vector Databases & Embeddings
- 4.2 Document Loading, Chunking & Preprocessing
- 4.3 RAG Pipeline Design & Implementation
- 4.4 Advanced RAG (Hybrid Search, Re-ranking)
- 054 labs
Fine-Tuning & Model Customization
Customize LLMs with transfer learning, LoRA/QLoRA, instruction tuning, RLHF concepts and model quantization.
- 5.1 Transfer Learning & Domain Adaptation
- 5.2 Parameter-Efficient Fine-Tuning (LoRA, QLoRA)
- 5.3 Instruction Tuning & RLHF Concepts
- 5.4 Model Quantization & Optimization
- 065 labs
AI Agents & Agentic Systems
Build AI agents: ReAct architecture, LangChain/LlamaIndex, function calling, multi-agent orchestration and memory management.
- 6.1 AI Agent Architecture (ReAct, Tool Use)
- 6.2 LangChain / LlamaIndex Frameworks
- 6.3 Function Calling & Tool Integration
- 6.4 Multi-Agent Systems (CrewAI, AutoGen)
- 6.5 Memory & State Management for Agents
- 075 labs
Multimodal Generative AI
Work with vision-language models, text-to-image/speech/video generation and document AI for structured extraction.
- 7.1 Vision-Language Models (CLIP, BLIP, LLaVA)
- 7.2 Text-to-Image (Stable Diffusion, DALL-E)
- 7.3 Text-to-Speech & Speech-to-Text (Whisper)
- 7.4 Text-to-Video & Video Understanding
- 7.5 Document AI (OCR + LLM)
- 085 labs
Gen AI Application Development
Build real applications: chat UIs, FastAPI services, evaluation benchmarks, guardrails and cost-optimization strategies.
- 8.1 Building Chat Applications (Streamlit, Gradio)
- 8.2 API Design for Gen AI Services (FastAPI)
- 8.3 Evaluation & Benchmarking (BLEU, ROUGE)
- 8.4 Guardrails, Content Filtering & Safety
- 8.5 Cost Optimization & Caching Strategies
- 095 labs
MLOps for Generative AI
Production operations: model serving (vLLM, Ollama), containers, monitoring with LangSmith, CI/CD and version control.
- 9.1 Model Serving & Deployment (vLLM, TGI)
- 9.2 Containerization & Cloud Deployment
- 9.3 Monitoring & Observability (LangSmith)
- 9.4 CI/CD for AI Applications
- 9.5 Version Control for Models & Prompts
- 105 labs
Responsible AI & Ethics
Build trustworthy AI: bias detection, hallucination pipelines, red teaming, data privacy and explainability tools.
- 10.1 Bias Detection & Mitigation
- 10.2 Hallucination Detection & Factuality
- 10.3 AI Safety, Alignment & Red Teaming
- 10.4 IP, Copyright & Data Privacy
- 10.5 Explainability & Transparency
- 115 labs
Domain-Specific Applications (Capstone)
Apply Gen AI to real domains: healthcare, finance, education, software engineering and creative-arts capstone projects.
- 11.1 Gen AI for Healthcare
- 11.2 Gen AI for Finance
- 11.3 Gen AI for Education
- 11.4 Gen AI for Software Engineering
- 11.5 Gen AI for Creative Arts
Eight stages, about two hours.
- 01
Reading
A short primer on the idea and why it matters.
- 02
The Hook
A quick demo that shows what you're about to build.
- 03
Concept Build
Interactive visuals that build intuition step by step.
- 04
Guided Build
Write the core code yourself, with hints at every step.
- 05
Exercises
Practice problems that check your understanding.
- 06
Challenge
An open-ended problem to solve on your own.
- 07
Reflect & Quiz
Lock in what you learned with a short quiz.
- 08
Reference Code
A complete, annotated Python implementation to study.
Real code, zero setup
Write and run Python right inside the lab. No installs, no environment headaches.
Intuition before equations
Interactive visuals let you poke at embeddings, attention and loss curves until the idea clicks.
Hints, not answers
Progressive hints go from a gentle nudge to a full walkthrough, so you get unstuck without being handed the solution.
Pick up where you left off
Every lab tracks the stages you've finished, so you can stop anytime and come back later.
Learning path
Each module builds on the last
- Foundations
- Core Gen AI
- NLP & LLMs
- RAG
- Fine-tuning & agents
- Apps & MLOps
- Capstone
Who it's for
Engineering students who want to build
Some Python helps, but you don't need prior AI experience. Module 1 covers the NumPy, neural-network and training-loop basics everything else builds on. Start there if you're new, or jump straight to the module you need.