SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 150,000 - 230,000 / annual
Clockwork Systems, founded by Stanford researchers and veteran systems engineers, is building software-driven fabrics to increase GPU cluster utilization. The company delivers cross-stack observability, workload fault tolerance, and performance acceleration for AI infrastructure by dynamically routing and pacing traffic to avoid congestion.
You will work as a Senior HPC Developer on cutting-edge GPU and high-performance networking problems. This is a hands-on role in a startup environment where you'll work across multi-node, multi-GPU systems and learn deeply across the full stack—from kernel and drivers to GPUs and networks.
Key responsibilities:
- Build and optimize high-performance GPU and networking subsystems
- Work with collective communication libraries and algorithms for multi-node, multi-GPU workloads
- Debug performance issues across kernel, driver, GPU, and network layers
- Develop and improve GPU-aware networking solutions
- Profile, analyze, and tune system performance using low-level tooling
- Collaborate closely with a small engineering team and take ownership of core systems
Requirements:
- 5+ years of experience in systems, HPC, or performance-critical software development
- Strong proficiency in low-level C/C++
- Solid understanding of RDMA networking, including InfiniBand, RoCE, and IBVerbs
- Experience working with multi-node, multi-GPU workloads
- Familiarity with collective communication libraries and communication algorithms
- Ability and willingness to debug complex issues across hardware and software boundaries
- Curiosity and eagerness to learn in a fast-moving startup environment
Bonus qualifications:
- Experience with congestion control mechanisms such as DCQCN
- Exposure to GPU-aware networking or advanced communication optimizations
- Experience with performance profiling, tracing, or observability tooling
- Background in AI infrastructure, HPC clusters, or distributed systems
