GenerativeAI and Prompt Engineering Training Program

Learning Outcomes:

Describe GenAI and differentiate from discriminative AI.

Explain fundamental concepts and applications of AI and GenAI in different domains.

Identify key GenAI models and their capabilities for text, code, and images.

Analyze the transformative role of GenAI in business operations and innovations.

Understand Transformer architecture, attention mechanism, tokenisation, context windows, and Model Context Protocol.

Apply fundamental and advanced prompt engineering techniques (Chain‑of‑Thought, Tree‑of‑Thought, self‑refinement).

Topics covered:

Introduction to Generative AI landscape.

Transformer architecture, attention, embeddings, tokenisation, context windows.

Model Context Protocol for context management.

Prompt engineering fundamentals: techniques, best practices, risks.

Advanced prompting: Chain‑of‑Thought, Tree‑of‑Thought, self‑criticism, structured output control.

Hands-on Lab Activities:

Set up cloud Python environment (Google Colab, Jupyter).

Interact with public GenAI models (ChatGPT, Gemini) for text generation.

Explore tokenisation and embeddings with pre‑trained models; experiment with context window sizes.

Craft prompts for summarisation, translation, content generation; test prompt injection attempts.

Implement Chain‑of‑Thought for multi‑step reasoning and generate structured JSON output.

Apply self‑refinement to improve factual accuracy.

Learning Outcomes:

Understand vector embeddings, vector databases, and similarity search types.

Explain RAG pipeline and its components; address LLM limitations.

Use LangChain core components (Models, Prompts, Chains, Memory, Tools) to orchestrate LLM applications.

Build simple and complex chains, and integrate external data with document loaders and text splitters.

Topics covered:

Vector embeddings and vector databases (ChromaDB, Weaviate).

Retrieval Augmented Generation (RAG) – pipeline, chunking strategies, retrieval quality.

LangChain: introduction, core components, LCEL, document loaders, text splitters.

Chains (simple, sequential, router) and memory integration.

Hands-on Lab Activities:

Generate embeddings using OpenAI Embeddings / Sentence‑Transformers.

Set up a local vector database and perform similarity searches.

Build a simple RAG pipeline from scratch; experiment with chunking strategies.

Implement QA system using RAG.

Create LangChain chains (simple, sequential, router) and integrate document loaders.

Learning Outcomes:

Implement multimodal LLMs (text, image, audio) and cross‑modal retrieval.

Apply fine‑tuning techniques (LoRA, QLoRA) for LLM customisation.

Design agentic AI systems: core components, architectures, multi‑agent coordination.

Build multi‑agent workflows with LangGraph, AutoGen, and n8n; implement agent‑to‑agent protocol.

Leverage AI pair programming and GenAI across the software development lifecycle.

Topics covered:

Multimodal LLMs (Gemini, GPT‑4V) – principles, applications, content generation.

Fine‑tuning vs RAG vs in‑context learning; LoRA, QLoRA.

Agentic AI: core components (Memory, Tools, Planning, Action), design patterns, multi‑agent systems.

LangGraph, AutoGen, n8n for agent workflows and orchestration.

AI pair programming (GitHub Copilot, CodeWhisperer) and GenAI in SDLC.

Hands-on Lab Activities:

Use multimodal LLM (Gemini, GPT‑4V) to analyse images and generate descriptions.

Fine‑tune a small open‑source LLM (e.g., LLaMA2‑7B) using LoRA.

Design a single‑agent system with tools (calculator) using LangGraph.

Develop a multi‑agent system where agents collaborate; implement agent‑to‑agent protocol.

Create n8n workflows to automate agent interactions.

Use AI tools for code generation, test case creation, and refactoring suggestions.