AMD AI Reinforcement Learning Hackathon & Workshops jointly hosted by AMD, Yardi School of AI, IIT Delhi and Unsloth.
Why Attend?
Join India’s leading AI engineers, researchers, and enthusiasts in Delhi for a 2-day hands-on workshops and hackathons powered by AMD’s MI300X GPUs. Whether you're a student, AI expert, or just starting your journey in machine learning, this is your chance to:
Learn hands-on techniques for Reinforcement Learning for AI models with Unsloth from Daniel Han, the creator of Unsloth
Fine-tune pre-trained models with tools like Unsloth to specialize them for real-world tasks
Work directly on AMD MI300X GPUs with guidance from AMD experts
Win exciting prizes in the hackathon
Seats are limited – register early to secure your spot in Delhi!
Looking for additional AMD AI Developer Benefits & Events? Sign up for the AI Developer Program to stay up to date and in the loop!
Agenda
Day 1:
9:00 AM – Doors Open- Registration & Check-in
09:45 - 10:05 AM – AMD Opening talk - Archana Vemulapalli - CVP, Global Enterprise Sales and AI Commercialization, AMD
10:05 - 10:25 AM – Guest talk by Dr. Sandeep Kumar, Assoc Professor in Electrical Engineering & the Yardi School of AI at IIT Delhi
10:30 - 10:45 AM - Hackathon Intro and Logistics Details, Daniel Wang, AMD
10:45 - 11:30 AM – Workshop 1: Teaching Models to Think: GRPO with Unsloth on a MI300X GPU, Daniel Han, CEO & Cofounder of Unsloth.ai
11:30 AM - 12:00 PM - Workshop 2: Finetuning AI models using Unsloth by Hitesh Kandala, AMD
12:00 PM – Lunch Break + Team Formation if not done already, Team registration
1:00 - 1:30 PM – Hackathon Overview Rules, Resources & Support, Aditya Kumar Singh & Maharshi Trivedi, AMD
1:30 PM - Hackathon Begins
6:30 PM Onwards – Dinner & Networking
Day 2
8:00 AM – Doors Open - Breakfast and Networking
9:30 AM – 1:30 PM - Hackathon/Competition Continues
1:00 PM – 3:00 PM - Lunch served
1:30 PM - 2:00 PM - Hackathon/competition Final Submission
2:00 PM – 4:00 PM - Evaluation: Prof. Srikanta Bedathur, Head, Amarnath & Sashi Khosla School of IT
3:00 PM - 3:30 PM - Guest talk by Souvik Chakraborty, Assoc Prof. Dept of Applied Mechanics at IIT Delhi
3:30 PM – 4:00 PM – AMD Talk - Dr Pratik Prabhanjan Brahma, Sr. Manager, AI Models Research, AMD
4:00 PM - 5:00 PM - Break
5:00 PM - 5:30 PM - Award Ceremony: Prof. Parag Singla, Head, Yardi School of AI
5:30 PM Onwards – Dinner
Workshop Details:
Workshops do not need any prior AI development experience.
Teaching Models to Think: GRPO with Unsloth on a MI300X GPU
Tutor: Daniel Han, CEO and Cofounder Unsloth.ai
Every major AI lab is using reinforcement learning to reach AGI. But why is reinforcement learning important? How does it actually work, and what are RL environments, reward functions, reward hacking, and GRPO? In this talk, we'll explore real-world use cases of reinforcement learning across different domains and see how you can experiment with it yourself, completely locally, by using Unsloth and AMD GPUs.
Fine-Tuning AI Models
Tutor: Hitesh Kandala, AMD Model R&D Team
This second beginner-level workshop focuses on fine-tuning pre-trained AI models to make them smarter or more specific to a task (like better in answering questions or generating images). You’ll use AMD GPUs and open-source tools like Unsloth, with guidance from AMD engineers to finetune the models.
Hackathon & Competition Details
A total of INR 5 Lakhs Prize money + GPU Cloud credits.
2X First prize: INR 1.25lakh + 500 hours GPU credits
2X Second prize: INR 75000 + 300 hours GPU credits
2X Third prize: INR 50000 + 200 hours GPU credits
The hackathon will have two tracks; each participating team will have to choose either of the tracks on the day of the event.
Track 1: AMD AI Premier League (AAIPL)
A head-to-head AI competition where teams of up to 4 developers build two intelligent language model based agents:
Q-agent: Generates valid, challenging multiple choice questions belonging to a given domain.
A-agent: Attempts to answer questions posed by the opposing team’s Q-agent.
Goal: Create a question generator that can pose the most difficult yet correct questions while ensuring your answerer can accurately answer as many of them.
Format:
Matches are played between pairs of teams where one team's Q-agent generates a set of questions to which A-agent of the opposing team responds and vice-versa.
Winning teams advance to the next stage.
Resources Provided:
1xMI300 GPU with 192G HBM for 24 hours
Sample code for prompt tuning, reinforcement learning, and fine-tuning
Deliverables:
Final working solution
Slides/ short video summarizing techniques used
Track 2:
Gaming the Models
A team of up to 4 developers will collaborate in developing a reinforcement learning pipeline for a given game environment. Each team will:
Design the input prompt for an LLM to output the strategy for playing and winning a game
Design reward functions around given constraints of the game
Decide on the reinforcement learning policy and hyperparameter around fine-tuning the LLM
Goal: Teach an LLM to win the game under various conditions using RL techniques. Team scoring highest points will win
Resources Provided:
1xMI300 GPU with 192G HBM for 24 hours
Sample code for prompt tuning, reinforcement learning, and fine-tuning
Deliverables:
Final working solution
Slides/ short video summarizing techniques used
Venue:
Lecture Hall Complex - Room Nos: LH-111 and LH-408,
Indian Institute of Technology Delhi, Hauz Khas, New Delhi, Delhi 110016, India
Please Read before Registering:
This hackathon is not beginner friendly. We are specifically looking for developers with experience and a proven track record in the following areas:
Proficiency in Python and expertise with ML libraries such as PyTorch, Hugging Face Transformers, or LangChain.
Prior experience (minimum 1 year) with training, fine-tuning, or prompting LLMs (e.g., Qwen, LLaMA, Mistral, or GPT-based models).
Ability to work with Linux-based remote GPU instances or cloud environments.
Based on the registration information provided AMD will approve the participants.
While it is good to register with a team, you will have an opportunity to form teams at the event.
IMPORTANT NOTE:
Please do not contact the event hosts for approval updates. All application approvals will be communicated only via Luma.
We are currently reviewing a high volume of applications. Thank you for your patience!
AMD is NOT responsible for sponsoring travel or accommodation. If your application is approved, it will be your responsibility to make your own travel and stay arrangements.
Please do NOT contact IIT, Delhi about event approval.