ORCA | Devpost

Inspiration

Training AI models traditionally requires access to expensive and centralized computational resources. This creates a barrier for smaller companies, startups, and independent developers who may not have the financial means to access high-performance GPUs for model training. At the same time, significant amounts of computing power in individual systems go underutilized, wasting potential resources.

What it does

Our platform offers a decentralized and cost-efficient solution for training AI models using a distributed network of idle GPUs. The decentralized AI training platform leverages blockchain and distributed computing to create a peer-to-peer network where participants can contribute their GPU resources to train AI models. This allows for more affordable and accessible AI development.

How we built it

The platform was built using a technical stack that includes React, Tailwind CSS, NEAR Protocol for blockchain integration, PyTorch Distributed for distributed training, and a serverless microservices architecture.

Challenges we ran into

Key challenges included synchronizing distributed training, ensuring data privacy, designing sustainable token economics, and maintaining network stability.

Accomplishments that we're proud of

We're proud of several accomplishments, including successfully implementing federated learning at scale, onboarding over 12 GPU contributors, and enabling 50+ research projects while reducing training costs by 60%.

What we learned

The project provided valuable technical insights into blockchain scaling solutions, distributed systems optimization, and federated learning implementation. It also offered business insights into token economy design, community building, and market positioning.

What's next for ORCA

The future roadmap includes implementing advanced scheduling algorithms, adding support for more AI model architectures, enhancing security features, expanding to new markets, and strengthening partnerships with research institutions and AI communities