Hands-On Generative AI

Foundations of AI and Machine Learning

• Machine learning vs rule-based systems
• Supervised and unsupervised learning paradigms
• Real-world enterprise use cases
• ML workflow overview:
  
  • Data preprocessing
  • Feature engineering
  • Overfitting and generalization
  • Model evaluation metrics

Hands-On:

• Training and evaluating a basic classification model

Deep Learning Primer

• Core deep learning concepts
• Neural network architectures and components
• Optimisation techniques and backpropagation
• Overview of popular deep learning frameworks

Hands-On:

• Image classification using a deep learning framework

Introduction to Generative AI

• What is Generative AI and why it matters
• Types of generative models and architectures
• Autoencoders and representation learning
• Variational Autoencoders (VAEs)

Hands-On:

• Training autoencoders and VAEs to generate synthetic data

NLP: Understanding Language as Data

• NLP fundamentals and applications
• Tokenization and vectorization techniques
• Embeddings and semantic representations

Hands-On:

• Finding similar documents using word embeddings

Large Language Models (LLMs)

• Evolution of NLP before and after LLMs
• Overview of transformer-based models
• Encoder and decoder-based architectures
• Practical applications of LLMs

Hands-On:

• Working with pre-trained language models
  

Language Generation & Prompt Engineering

• Common generative tasks:
  
  • Text completion
  • Summarization
  • Dialogue systems
  • Code generation
    
• Prompt engineering principles
• Prompt refinement techniques

Hands-On:

• Prompting exercises for summarization, code generation, and text labeling

Adapting Pre-trained Models

• Transfer learning and fine-tuning strategies
• Cost and performance considerations
• Catastrophic forgetting
• Sampling techniques for controlling model output

Hands-On:

• Fine-tuning a language model for sentiment analysis
• Controlling output using temperature, Top-K, Top-P, and beam search
  

Retrieval-Augmented Generation (RAG)

• Limitations of standalone LLMs
• RAG architecture and workflow
• Integrating retrieval with generative models
• Optimisation and deployment considerations

Capstone Project

• Building a conversational system using Retrieval-Augmented Generation
• Applying learned techniques to a real-world dataset
• Evaluating model performance and behaviour