Product Science provides a decentralized AI training platform that orchestrates heterogeneous compute resources worldwide into a trustless, fault‑tolerant environment. By using open protocols for coordination and resource allocation, it enables elastic scaling across GPUs and AI ASICs without reliance on centralized data centers, lowering entry barriers for frontier‑scale model training.
Funding
$18M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
9OFounders
Product
Problem
Frontier AI model training is constrained by the high cost and limited availability of specialized GPU hardware, which is concentrated in a few large data centers. Distributed compute across heterogeneous, geographically dispersed resources introduces significant engineering challenges such as communication overhead, fault tolerance, and the need for trusted coordination.
Solution
Product Science builds a decentralized AI training platform that orchestrates compute resources worldwide into a unified, trustless training environment. By defining open protocols for coordination, fault tolerance, and resource allocation, the system enables elastic scaling across general‑purpose GPUs and specialized ASICs without reliance on a single data center. The architecture treats node instability as an inherent property, providing built‑in resilience and permissionless access for participants. This approach lowers entry barriers, allows organizations to meet data‑sovereignty and IP compliance requirements, and creates a market‑driven pricing model for frontier‑scale AI workloads.
Target Audience
Primary customers are AI research labs, enterprise machine‑learning teams, and independent developers who require frontier‑scale training but lack access to large centralized GPU clusters.
Features
- Open, permissionless protocol for end‑to‑end orchestration of heterogeneous compute nodes
- Fault‑tolerant scheduling that automatically handles unstable or intermittent participants
- Communication layer optimized for low‑overhead data exchange across geo‑distributed hardware
- Support for both general‑purpose GPUs and specialized AI ASICs within the same training job
- Incentive mechanisms that let hardware providers monetize idle capacity while keeping pricing market‑driven
- Compatibility layer for existing AI frameworks, enabling seamless migration of workloads to the decentralized network