Workshop: Master Fine-Tuning LLMs — From Weights to Wisdom

Minimum 2 years of experience. Freshers and college students will not be permitted, so please excuse this event.

Lunch is on your own, will not be provided

## Workshop Overview

Large Language Models (LLMs) are powerful, but out of the box they’re generic.
To make them truly useful in your domain, you need to fine-tune them.
This hands-on workshop will take you beyond prompt engineering into the world where you retrain and adapt models by updating weights — the real deal.

### What You’ll Learn

We’ll cover all three approaches to fine-tuning:

1. Full Fine-Tuning – Updating all weights of the model for maximum customization.

• When to use it (and when it’s overkill).
• GPU/compute requirements.
• Walkthrough: fine-tuning a GPT-like model on a custom dataset.

2. Parameter-Efficient Fine-Tuning (PEFT) – LoRA, QLoRA, adapters.

• Update only a small subset of weights (millions vs billions).
• Train on a laptop or low-cost GPU.
• Demo: LoRA fine-tuning for domain-specific chatbots.

3. Instruction Fine-Tuning & Alignment (SFT + RLHF) –

• Teach models to follow instructions, tone, and style.
• Use reward models and human feedback loops.
• Hands-on: building a mini RLHF pipeline.

### Who Should Attend

• ML Engineers & Data Scientists who want to go beyond prompt hacks.
• Developers exploring custom AI assistants for their company or product.
• Researchers curious about alignment & efficiency tricks.
• Startup founders wanting to differentiate their AI with domain-trained LLMs.

### Takeaways

• Hands-on code notebooks you can reuse.

• Understanding of trade-offs between the 3 approaches.

• A fine-tuned mini-LLM you can deploy.

• Confidence to choose the right method for your projects.

• ⚡ Seats are limited — this workshop is designed to be interactive and practical.

Sign up now and start bending LLMs to your will.

***

## Prerequisites

• Basic Python knowledge.
• Laptop with Colab notebook access.
• Curiosity and enthusiasm (we'll teach you everything else from the basics!)