Teams at Clockwork Systems, Inc.

Design and implement low-level systems software for fault-tolerant, high-performance distributed GPU training. Modify GPU frameworks and runtimes, build reliable large-scale training components, debug complex concurrent and distributed systems, and own solutions from design through production. The role requires strong systems programming, concurrency and memory-model expertise, distributed-systems reasoning, and the ability to navigate large codebases. GPU programming, high-performance networking, ML runtimes, and cluster orchestration are preferred.

Build and own customer-facing installation, configuration, observability, documentation, and deployment capabilities for TorchPass. Debug production infrastructure, support customer evaluations, distinguish product defects from environment issues, improve tests and documentation based on field feedback, define cluster requirements and guarantees, and extend the product to new schedulers and job launchers. The role requires strong Python, Go, Kubernetes or Slurm, infrastructure operations, customer support, and technical documentation skills.

Develop and optimize high-performance GPU and networking subsystems for multi-node, multi-GPU workloads. Build GPU-aware networking solutions, work with collective communication libraries, and debug performance issues across kernel, driver, GPU, and network layers. Profile and tune systems using low-level tools while collaborating with a small engineering team on distributed computing infrastructure.