Course Details

• Basic knowledge of Python programming

• Familiarity with machine learning fundamentals

• Open to All.

• Undergraduate/ Postgraduate in Computer Science, Artificial Intelligence, Data Science, or related disciplines (completed/pursuing) and even working professionals in similar domains.

This course is designed to

• Provide a foundational and practical understanding of Generative AI and its implementation across multiple domains.

• Understand the architecture and workflow of large language models, diffusion models, and generative pipelines.

• Enable students to do hands-on model building, fine-tuning, and deployment to solve industry-grade generative AI problems.

The key modules of the course consist of:

• Creating the high level implementation architecture.

• Intro to Transformer Architecture.

• Teacher Forcing & Masked Attention & Zooming into Decoder Layer.

• Positional & Sinusoidal Encoding.

• Batch Normalization & Layer Normalization.

• Introduction to Language Modeling, Causal Language Modeling, Pre-Training & Fine-Tuning.

• Decoding Strategies: Exhaustive, Greedy, Beam Search.

• Sampling Strategies – Top P & Top K.

• BERT-MLM – Next Sentence Prediction.

• Pre-training Parameters & Adapting to Downstream Tasks.

• Tokenization Challenges, Subword Tokenization.

• Byte Pair Encoding, WordPiece Tokenizer, SentencePiece Tokenizer.

• Intro to BART and GPT Models.

• Fine-tuning Strategies & Multi-task Learning.

• Intro to RAG.

• Different Types of RAG Techniques

• Text-to-text and text-to-image generation models.

• Open-source generative AI frameworks (Hugging Face, LangChain, Diffusers.

• Building generative AI pipelines for enterprise applications.

• Deployment strategies and serving models at scale.

• Ethics, bias, and safety in generative AI.

• Capstone project.

The milestones are:

• Build a Rag application end to end.

• Play with various vector DB’s, retrieval techniques etc etc.

• Finetune a small model for small task.

• Know how to connect the front end and the backend

• Basics of deploying it in the cloud

• Weightage for assignments = 20%

• Weightage for Quiz = 20%

• Weightage for live session attendance = 10%

• Weightage for project = 50%

At the end of this course, students should be able to:

• Evaluate, optimize, and deploy generative models for production use cases

• Build and fine-tune generative AI models for various applications