Generative AI & Agentic AI with Python
What You’ll Learn
Using ChatGPT is one thing, but a real AI engineer is someone who can look inside LLMs, fine-tune them, and build production AI apps. This course teaches you exactly that — through Transformers, RAG, LangChain, LangGraph, and Multi-Agent Systems with 10+ AI projects.
Benefits
- LLM Fundamentals — Transformer Architecture
- Prompt Engineering — CoT, ReAct, ToT
- LLM Fine-tuning — LoRA, QLoRA, PEFT
- RAG with ChromaDB, Pinecone, FAISS
- LangChain — Chains, Agents, Memory, Tools
- LangGraph — Stateful Agentic Workflows
- CrewAI — Multi-Agent Orchestration
- 10+ Real AI Application Projects
Curriculum
💡 Module 1 — Foundations of Generative AI
- History of AI: From RNNs to Transformers
- Transformer Architecture: Self-Attention, Multi-Head Attention, Positional Encoding
- BERT: Masked Language Modeling, NSP, Sentence Embeddings
- GPT Series: Autoregressive Generation, GPT-2, GPT-3, GPT-4 internals
- T5, LLaMA, Mistral, Gemma, Phi — comparative study
- Tokenization: BPE, WordPiece, SentencePiece
🎯 Module 2 — Prompt Engineering Mastery
- Zero-shot, One-shot, Few-shot Prompting
- Chain-of-Thought (CoT), Tree-of-Thought (ToT)
- ReAct Pattern: Reasoning + Acting for AI agents
- Self-consistency, Generated Knowledge Prompting
- System Prompts, Role Prompting, Persona Design
- Output Formatting: JSON mode, Structured Outputs, Guardrails
🔧 Module 3 — LLM Fine-tuning
- When to Fine-tune vs RAG vs Prompt Engineering
- Dataset Preparation: JSONL format, Instruction tuning datasets
- LoRA (Low-Rank Adaptation): Math, Configuration, Training
- QLoRA: 4-bit Quantization + LoRA on consumer GPUs
- PEFT Library: Training with HuggingFace Transformers
- RLHF: Reward Modeling, PPO, DPO (Direct Preference Optimization)
- Model Evaluation: ROUGE, BLEU, BERTScore, LLM-as-Judge
📚 Module 4 — RAG Systems
- RAG Architecture: Indexing, Retrieval, Augmentation, Generation
- Embedding Models: OpenAI, Cohere, HuggingFace sentence transformers
- Vector Databases: ChromaDB (local), Pinecone (cloud), FAISS
- Chunking Strategies: Fixed, Recursive, Semantic, Parent-Child
- Advanced Retrieval: Hybrid BM25+Dense, Cross-encoder re-ranking
- Evaluation: Faithfulness, Relevance, Context Recall (RAGAs)
- Projects: PDF Q&A Bot, YouTube Video Chatbot, Legal Document Analyzer
⚙️ Module 5 — LangChain & LangGraph
- LangChain: LLMs, PromptTemplates, Chains, Memory, OutputParsers
- LangChain Tools: Tavily Search, Wikipedia, Calculator, Custom Tools
- LCEL (LangChain Expression Language): Pipe operator, Runnables
- LangGraph: StateGraph, Nodes, Edges, Checkpointing, Streaming
- Agent Patterns: ReAct Agent, Reflection, Corrective Loops
🤖 Module 6 — Multi-Agent Systems & Deployment
- CrewAI: Agents, Tasks, Roles, Backstory, Tools, Crew execution
- Multi-Agent Coordination: Task delegation, Agent communication
- FastAPI: Building REST APIs for AI applications
- Streamlit & Gradio: Rapid AI app UIs
- Docker + AWS EC2: Deploying AI apps to production
