Customizing Generative AI Models 

Duration: 3 days / 24 Hrs

Introduction 

This 24-hour, hands-on course focuses on customizing Generative AI models to behave the way your applications need—using prompt engineering, RAG, parameter-efficient fine-tuning, and agentic workflows. The emphasis is on practical customization techniques rather than black-box model usage. 

Objectives 

By the end of this course, participants will be able to: 

  • Understand how LLMs work and where customization fits 
  • Customize model behavior using prompts, RAG, and PEFT techniques 
  • Evaluate and benchmark GenAI models using task-specific metrics 
  • Build scalable, agentic AI workflows using LangGraph 

Key Takeaways 

Participants will leave with: 

  • A practical toolkit for customizing LLM behavior without full retraining 
  • Hands-on experience with RAG and parameter-efficient fine-tuning 
  • Confidence to choose the right customization strategy for real use cases 

Training Methodology (Learning by Doing) 

  • Hands-on labs throughout—every concept is applied immediately 
  • Incremental builds from prompt tuning to fine-tuned and agentic systems 
  • Realistic datasets and scenarios instead of toy examples 
  • Experiment, measure, refine using evaluation-driven development 

Course Outline: 

Understanding Generative AI 

  • Definitions: Intelligence, AI, Generative AI 
  • Differences from Traditional AI 
  • Tokenization and how GenAI works 
  • Benefits and challenges 
  • Popular models and frameworks 
  • LLM settings and parameters 

Prompt Engineering 

  • Direct prompting (zero-shot) 
  • One-shot and few-shot examples 
  • Chain-of-Thought and Tree-of-Thoughts 
  • Common mistakes and refinement techniques 

Building LLM-Based Applications 

  • Design building blocks 
  • Accessing LLMs via APIs 
  • Prompt templates 
  • Conversational completion models 
  • Batch APIs and cost control 

Evaluating Generative AI Models 

  • GenAI vs Predictive AI evaluation 
  • Metrics and benchmarking 
  • Custom criteria and chat-specific metrics 

Prompt Engineering for Customization 

  • Theory and behavior modification 
  • Automatic and dynamic prompt generation 
  • Troubleshooting and refinement 

Retrieval Augmented Generation (RAG) 

  • Embeddings and indexing 
  • Retrieval techniques: vector, full-text, fusion 
  • Filtering, reranking, looping 
  • Contextual generation 

Parameter-Efficient Fine-Tuning (PEFT) 

  • Comparison with traditional fine-tuning 
  • Prompt-based and low-rank adaptation techniques 
  • IA3 and other methods 
  • Synthetic data generation and evaluation 
  • Bias and data balance considerations 

Agentic AI with LangGraph 

  • LangGraph principles 
  • Multi-agent workflows 
  • Communication, coordination, error handling 
  • Scaling agentic applications