Week 1 – Introduction to Generative AI

• Evolution: Rule based -> ML -> DL -> Foundation Models
• What makes LLMs powerful?
• Overview of some popular GenAI tools (GPT, Gemini, Claude, Llama)
• Open-Source Vs Closed models
• Real Industry use cases across domains
  
  Hands-on
• Using an LLM via API (OpenAI / Azure / open-source via Hugging Face)
• Simple prompt experiments (Summarization, Classification & Content generation

Week 2 – NLP & Deep Learning Essentials

• Tokens, embeddings, cosine similarity
• Language modeling intuition
• Why word order matters
• Encoder vs Decoder models
• Attention mechanism (high-level intuition)
  
  Hands-on
• Generate embeddings
• Visualize semantic similarity
• Build a simple semantic search

Week 3 – Transformers and LLM Internals

• Transformer architecture (self-attention, layers)
• Context window & token limits
• Training vs inference
• Hallucinations: why they happen
• Temperature, top-p, max tokens
  
  Hands-on
• Experiment with decoding parameters
• Compare outputs across different LLMs
• Failure analysis of hallucinations

Week 4 – Prompt Engineering

• Prompting is programming
• Prompt patterns: Zero-shot, Few-shot, Chain-of-Thought, ReAct, Self-consistency
• Prompt injection & security risks

  Hands-on

• Design prompts for: Structured outputs (JSON), Reasoning tasks, Role-based behavior
  
• Build a prompt library

Week 5 – Retrieval Augmented Generation (RAG)

• Why LLMs fail on private data
• RAG architecture
• Advanced RAG patterns (Hybrid search (keyword + vector), Re-ranking, Metadata filtering, Multi-hop RAG)
• Chunking strategies
• Vector databases (FAISS, Pinecone, Chroma)
• Embedding models trade-offs

  Hands-on

• Build an end-to-end RAG system: Ingest documents, Create embeddings & Query + generate answers
• Test hallucination reduction
• Naive RAG vs optimized RAG

Week 6 – Fine Tuning & Customization

• Prompting vs Fine-tuning vs RAG
• When NOT to fine-tune
• When RAG beats fine-tuning (real examples)
• Instruction tuning
• LoRA & PEFT concepts

  Hands-on

• Fine-tune a small open-source LLM
• Compare base vs fine-tuned performance
• Cost & performance analysis

Week 7 – Evaluation, Guardrails & Reliability

• Why GenAI evaluation is hard
• Automatic vs human evaluation
• Metrics: Faithfulness,, Relevance &Toxity
• Guardrails & moderation

  Hands-on

• Build an evaluation pipeline
• Add: Output validation, Safety checks
• Use LLMs to evaluate LLMs (LLM-as-judge

Week 8 – Agents

• What are LLM agents?
• Planning, memory, tools
• ReAct & function calling
• Single vs multi-agent systems

  Hands-on

• Build an agent that: Uses tools (search, calculator, APIs) & Maintains memory

Week 9 – Multimodal Generative AI

• Text-to-Image (Diffusion models)
• Image-to-Text
• Speech-to-Text & Text-to-Speech
• Video generation overview
• Invoice & contract processing
• KYC & compliance document review
• Medical reports / legal discovery
• Manufacturing defect analysis

  Hands-on

• Build: Image captioning app, Document

Combine text + image inputs

Week 10 – Scaling, Deployment & LLMOps

• Inference optimization
• Caching & batching
• Model serving architectures
• Cost optimization
• Monitoring drift & misuse
  
  Hands-on
• Deploy a GenAI app (API or UI)
• Add: Logging, Feedback loop
• Cloud deployment demo

Week 11 – Ethics, Privacy & Responsible AI

• Bias & fairness in GenAI
• Data privacy concerns
• Copyright & IP risks
• Regulatory landscape (EU AI Act, etc.)
• Enterprise governance models
  

Week 12 – Capstone Project & Industry Applications

Capstone Options:
Participants build one of:

• Enterprise chatbot
• Intelligent document processing system
• GenAI-powered analytics assistant
• Domain-specific advisor (banking, pharma, HR)
  
  Final Deliverables
• Architecture diagram
• Live demo
• Cost & risk assessment
• Business value articulation
  

  
  Tools & Platforms Covered
  Hugging Face, LangChain, LlamaIndex, Chroma, OpenAI API, Streamlit/Gradio, PEFT, GitHub