Part 1: Introduction to LLMs

• Evolution of NLP → Transformers → LLMs
• Use cases: Chatbots, Tutors, Coders, Analysts
• Overview of popular LLMs (Open & Closed source)
• Limitations & challenges of LLMs

Part 2: Python Foundations for LLMs

• Python environment setup (conda, venv)
• Working with APIs and JSON
• Text preprocessing in Python
• Using notebooks for experimentation

Part 3: Transformers & LLM Architecture

• Tokenization and embeddings
• Attention & self-attention
• Encoder–Decoder vs Decoder-only models
• Context window & scaling laws

Part 4: Prompt Engineering

• Prompt structures & templates
• Zero-shot, few-shot, chain-of-thought
• Role-based prompting
• Prompt testing & optimization

Part 5: LLM Integration Using Python

• Using Hugging Face Transformers
• Using OpenAI-compatible Python SDKs
• Streaming responses
• Cost and latency optimization

Part 6: Embeddings & Vector Databases

• Semantic search concepts
• Embedding generation in Python
• Vector stores (FAISS, Chroma, Pinecone – concept & hands-on)
• Similarity search & clustering

Part 7: Retrieval Augmented Generation (RAG)

• Why RAG is needed
• Document loaders & chunking strategies
• Building Q&A systems over PDFs, Docs, Websites
• Reducing hallucinations

Part 8: Tool Calling & Function Execution

• Concept of tool-enabled LLMs
• Designing Python functions for LLMs
• API & database integration
• Autonomous workflows

Part 9: Fine-Tuning & Model Adaptation

• When to fine-tune vs prompt
• PEFT, LoRA basics
• Dataset preparation
• Evaluation after fine-tuning

Part 10: Evaluation & Monitoring

• Accuracy, relevance & faithfulness
• Human vs automated evaluation
• Logging, tracing & observability
• Cost tracking

Part 11: Safety, Ethics & Governance

• Bias, hallucination & misuse
• Guardrails and content moderation
• Data privacy & compliance
• Responsible AI frameworks

Part 12: Deployment & Scaling

• Deploying LLM apps using FastAPI
• Cloud deployment concepts
• Caching & performance optimization
• CI/CD for AI applications

Capstone Project (Mandatory)

Participants will build Project using Python:

• AI Tutor for Academic / Competitive Exams

Tools & Technologies Covered

• Python
• Hugging Face
• LangChain / LlamaIndex
• Vector Databases (FAISS, Chroma)
• FastAPI
• Git & Cloud Basics