Decentrify

Decentrify

Inspiration

Machine learning today has an identity problem. Only people with strong ML backgrounds and access to expensive hardware can realistically train models. This leaves out subject-matter experts who have valuable data and ideas but lack ML expertise or compute resources. We asked ourselves: Why should hardware and technical background define who gets to do machine learning? That question inspired Decentrify.

What it does

Decentrify is a decentralized platform that lets anyone identify as a Machine Learning Engineer. Users upload their datasets and experiment goals without needing ML expertise or cloud infrastructure. Other users join as peers, contributing their available CPU or GPU compute from their own devices. Together, they form a decentralized training network that runs machine learning jobs collaboratively. For example, a PhD biologist with zero ML experience can upload biological data and train a compute-intensive model, while peers contribute their computing power to support the experiment. What makes Decentrify unique is that we abstract away machine learning complexity while enabling real compute sharing. We are building an “Uber for compute”, where users share CPU and GPU resources to power real experiments. This redefines identity in machine learning. You don’t need to be an ML engineer to be a researcher. You don’t need expensive cloud providers to run big experiments. Your identity becomes your idea, not your hardware or technical background.   

How we built it

We designed Decentrify around a coordinator-peer architecture. The coordinator uploads data and experiment parameters, while peers connect to contribute compute and participate in training. The system visualizes training progress, peer contributions, and model outputs in real time, making collaboration transparent and accessible even to non-technical users.

Challenges we ran into

Designing a system that abstracts ML complexity without limiting flexibility Coordinating distributed compute across multiple peers Balancing usability for non-technical users with technical depth Managing compute reliability and synchronization in a decentralized setting

Accomplishments that we're proud of

Creating a working decentralized compute-sharing concept Making machine learning accessible to non-technical users Designing an identity-driven ML platform centered on collaboration Building a system with real-world utility beyond experimentation

What we learned

We learned that accessibility in AI isn’t just about tools — it’s about identity. When barriers are removed, people from diverse backgrounds are empowered to experiment, learn, and contribute meaningfully to machine learning.

What's next for Decentrify

Incentivizing peer participation and compute sharing Improving model orchestration and fault tolerance Expanding support for real-world datasets and workloads Building stronger community identity and reputation systems