Agentic AI, Generative AI

---

 Goal:

• Understand the concepts of Generative AI, Foundation Models, Transformers, and Agentic AI architectures.
• Study how large language models (LLMs) work internally, including attention, embeddings, and tokenization.
• Learn agent architectures, multi-agent systems, tool-use, planning, and autonomous workflows.
• Use and fine-tune LLMs (LLaMA, GPT, Mistral).
• Build AI agents capable of reasoning, planning, tool-use, memory, and environment interaction.
• Create generative AI applications for text, images, audio, video, and multimodal tasks.
• Integrate LLMs with databases, APIs, RPA, and external tools.

• What is Generative AI?
• What is Agentic AI?
• AI vs LLMs vs Agents
• Foundation Models overview
• Capabilities and limitations of LLMs
• Real-world AI Agent use-cases

• Python refresher for AI
• Working with APIs (OpenAI, HuggingFace)
• Introduction to LLMs
• Tokenization, embeddings, prompt structure
• Prompt engineering (basic → advanced)
• Safety, hallucinations, guardrails

• Transformer architecture
• Multi-head attention
• Positional encoding
• Pre-training vs fine-tuning
• Open-source LLMs (LLaMA, Mistral)
• Model evaluation (BLEU, Rouge, perplexity)

• Text generation basic
• Story generation
• Summarization
• Question answering
• Retrieval-Augmented Generation (RAG)
• Knowledge base creation using embeddings & vector databases
• Prompt templates & chaining

 Images

• Diffusion models
• Stable Diffusion
• ControlNet
• Prompting techniques (style, composition, lighting)

 Audio

• Speech-to-text (Whisper)
• Text-to-speech
• Music generation basics

 Video

• Video generation tools
• AI animation and editing workflows

• Fine-tuning vs LoRA vs QLoRA
• Dataset preparation
• Safety tuning
• Reinforcement Learning from Human Feedback (RLHF) – Intro
• Model compression & optimization

• What is an AI Agent?
• Architecture of agents
• ReAct (Reasoning + Acting) framework
• Tool use: APIs, calculators, web search, databases
• Planning algorithms
• Memory systems: short-term vs long-term
• Multi-agent collaboration

• Autonomous agents (e.g., AutoGPT, BabyAGI concepts)
• Multi-agent systems
• Agent orchestration frameworks
• Workflow automation
• Task decomposition
• Real-time agent environments

• API deployment (FastAPI, Flask)
• Containerization (Docker)
• Frameworks: LangChain, LlamaIndex, Haystack
• Vector databases (Pinecone, Weaviate, FAISS)
• Cloud deployment (AWS, Azure, GCP)
• Monitoring, logging, versioning

• Practical
• Interview Preparation