Airpool | Devpost

The landing page
Showing Backend findings on the UI/UX
Conversation between different bots with each other
Our Story

What if every startup could harness the power of 1,000 Nvidia RTX 4090 GPUs or 500 AMD Ryzen 9 7950X CPUs, at a fraction of the cost? While big tech companies can afford to spend millions of dollars on their GPU infrastructure, small developers and startups are left struggling to access affordable computational power needed for innovation. This bottleneck stifles creativity and limits competition in an industry ripe for disruption. Our solution, AirPool, is an open-source decentralized network that democratizes access to high-performance computing. By allowing individuals and organizations to contribute their unused GPU and CPU cores, we create a shared, affordable infrastructure for distributing computational services. What's unique about AirPool is that resource availability is dynamically allocated based on project popularity, similar to GitHub repository stars. This ensures that groundbreaking open-source projects get the resources they need to thrive. AirPool empowers small companies and developers to tackle sophisticated AI models, train large language models, and perform complex data analysis without the heavy upfront investment in hardware. With our platform, computational resources become as accessible and affordable as open-source software. Join us in revolutionizing the computational landscape, making high-performance computing affordable, and leveling the playing field for innovators everywhere. With AirPool, we're not just sharing resources; we're fostering an open-source ecosystem that accelerates innovation for all.

Built With

Submitted to

Cal Hacks 11.0

Created by

As the lead backend developer for AirPool, I architected and implemented a multi-agent system using the Fetch.ai uAgents framework. My key contributions include:

1. Designed and implemented a decentralized resource allocation system consisting of three main components:
- A Project Lead Agent that defines computational requirements and manages resource requests
- Multiple Contributor Agents that offer their computational resources (GPUs and memory)
- A Dynamic Moderator Agent that optimizes resource allocation based on availability and project needs

2. Developed an efficient resource sharing mechanism that:
- Generates and manages random computational resources for contributor agents
- Maintains a shared resource pool using JSON file-based persistence
- Implements real-time resource allocation and optimization algorithms

3. Created a robust decision-making system that:
- Analyzes available computational resources
- Makes intelligent allocation decisions based on project requirements
- Generates detailed allocation reports in JSON format for frontend integration

4. Implemented scalability features allowing the system to handle multiple contributor agents (currently supporting 100 agents) with the flexibility to scale up based on demand.

The backend system successfully demonstrates the feasibility of creating a decentralized computational resource sharing network, laying the groundwork for democratizing access to high-performance computing resources.

Rishabh Agarwal
Manan Gulati
SaurSum8
Raghav Bansal

Updates

Rishabh Agarwal started this project — Oct 20, 2024 01:26 PM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.