Program Overview

Program Schedule & Modules

A comprehensive six-day curriculum designed to take you from LLM fundamentals to production-grade deployment

Program Highlights

  • 40+ hours of intensive instruction combining theory and GPU practice
  • 6 expert plenary speakers presenting cutting-edge research in LLMs and CI
  • Hands-on labs with PyTorch, HuggingFace, PEFT, TRL, and vLLM
  • Capstone projects with direct mentorship from industry instructors
  • Networking opportunities with peers, faculty, and industry leaders
  • Official IEEE CIS certificate upon successful completion

Daily Structure

09:00 - 10:45

Plenary presentation or keynote lecture

10:45 - 11:00

Coffee break & networking

11:00 - 13:00

Hands-on GPU lab session

13:00 - 14:00

Lunch break

14:00 - 16:00

Practical workshop / lab deep dive

16:00 - 17:30

Interactive discussion, Q&A, and clinic

What's Included

  • Complete notebooks, slides, and code repositories
  • Daily lunch and refreshments
  • GPU cloud compute access for training and inference
  • Networking events and banquet dinner
  • Direct instructor mentorship during labs
  • Official IEEE CIS Certificate of Completion
Curriculum

6-Day Program Schedule

Carefully designed curriculum with expert instruction, hands-on labs, and networking opportunities

Day 1

Monday

June 22, 2026

Foundations of LLM Fine-Tuning

08:30

Opening Ceremony & Welcome

Prof. Pallavi Nikumbh
Discussion

Meet the instructors and fellow participants. Program overview and logistics.

45 minutes
09:15

Plenary: Privacy & Security in LLM Systems

Dr. Debmalya Biswas
Plenary

Understanding big data analytics and security considerations in modern AI.

90 minutes
10:45

Break

Break
30 minutes
11:15

Lecture: Transformer Architecture Review & Pretraining Objectives

Dr. Kishor Burchundi
Lecture

Deep dive into transformer architectures and foundation model training.

120 minutes
13:15

Lunch

Lunch
60 minutes
14:15

Hands-on Lab: Setting Up Fine-Tuning Environment

Lab Instructors
Lab

Install libraries (HuggingFace, PEFT, bitsandbytes). Configure GPU workstations.

120 minutes
16:30

Break

Break
30 minutes
17:00

Discussion & Reflection

Dr. Anuradha Thakre
Reflection

Q&A, group discussion, and day wrap-up.

90 minutes

Learning Outcomes

Outcome 1

Understand transformer architectures, attention mechanisms, and scaling laws

Outcome 2

Master parameter-efficient fine-tuning (LoRA, QLoRA) on custom datasets

Outcome 3

Implement preference alignment with Direct Preference Optimization (DPO)

Outcome 4

Apply post-training quantization (GPTQ, AWQ, GGUF) without quality loss

Outcome 5

Deploy high-throughput inference engines with continuous batching via vLLM

Outcome 6

Design end-to-end production RAG pipelines and evaluate generation quality

Prerequisites

Required Knowledge

  • •Proficiency in Python programming and NumPy / Pandas
  • •Basic understanding of deep learning and PyTorch tensors
  • •Familiarity with neural network training (loss, backpropagation)

Technical Requirements

  • •Laptop with modern browser and high-speed Wi-Fi capability
  • •Google Colab / Kaggle account or modern terminal access
  • •All lab code and GPU runtimes will be provided

Note: Refresher notebooks and environment setup guides will be shared with all registered participants prior to the school start date.