Clockwork, founded in 2018, develops Software-Driven AI Fabrics™ - a programmable software layer designed to make large-scale AI clusters observable, resilient, and optimized for peak performance. The company's technology addresses core challenges in AI infrastructure, including GPU cluster optimization, network observability, fault tolerance, and dynamic traffic control. Its solutions are deployed across on-premises data centres, public clouds, and hybrid environments, serving sectors such as financial trading, high-tech, and online gaming.
Two products form the core of Clockwork's offering. FleetIQ is a software control plane that unifies nanosecond-level visibility, dynamic traffic control, and job-aware resilience. TorchPass is a workload fault-tolerance framework that enables live workload migration and path failover to keep AI training jobs running through hardware and network failures. According to SemiAnalysis, TorchPass is the only fault-tolerance framework that achieves this without compromising training performance.
The company's approach targets the growing complexity of operating large GPU clusters for AI workloads, where even brief interruptions can translate into significant compute waste. By providing a software layer that sits between hardware and applications, Clockwork aims to give infrastructure operators granular control over cluster behaviour without requiring changes to the underlying network fabric.
