About Clockwork Systems
Clockwork.io – Software Driven Fabrics to increase GPU cluster utilization
Clockwork Systems was founded by Stanford researchers and veteran systems engineers who share a vision for redefining the foundations of distributed computing. As AI workloads grow increasingly complex, traditional infrastructure struggles to meet the demands of performance, reliability, and precise coordination. Clockwork is pioneering a software-driven approach to AI fabrics by delivering cross-stack observability to catch and quickly resolve problems, workload fault tolerance to keep jobs running through failures, and performance acceleration that dynamically routes and paces traffic to avoid congestion.
About the Role
TorchPass keeps distributed PyTorch training running through hardware and network failures. The Product Engineering team makes it something customers can install, trust, and run on their own clusters.
We are looking for an experienced engineer who has turned working infrastructure software into a product that other people can deploy and operate. You will own the parts of TorchPass that customers touch, and you will be one of the engineers they hear from when something does not work.
This is a product and deployment role. It does not require GPU runtime, NCCL, or kernel experience.
What You Will Do
- Build and own the customer-facing surfaces of TorchPass: installation, configuration, and the views that show customers what the product is doing
- Keep product documentation accurate with every release
- Work directly with customers and our solutions engineers during evaluations: read logs, separate product defects from environment issues, and say clearly what happened
- Turn what we learn in the field into fixes, tests, and documentation
- Define, together with the core engineering team, what TorchPass needs from a customer's cluster and what it guarantees in return
- Extend TorchPass to new schedulers and job launchers as customers need them
What We're Looking For
- 4+ years building or operating infrastructure software in production
- You have taken a system that worked for its authors and made it installable and supportable by people who did not build it
- Strong debugging skills on production systems you did not write
