2 months ago
Remote, WorldwideSenior / Staff+
H1B Sponsor
Responsibilities
- Design, develop, and maintain comprehensive benchmarking frameworks spanning OS, kernel, and application layers.
- Profile workloads across CPU, memory, I/O, network, and accelerator subsystems to identify bottlenecks and optimization opportunities.
- Establish and own performance baselines across CIQ's product and solutions portfolio.
- Build and maintain automated performance regression-detection pipelines integrated into CI/CD workflows using Fuzzball.
- Identify, triage, and resolve regressions across user space, kernel space, and application layers.
- Drive proactive performance improvements across every layer of the stack.
- Optimize for AI inference and training workloads, including LLM serving and model parallelism.
- Develop deep expertise in CIQ's Fuzzball platform and integrate performance benchmarks into its pipelines.
- Collaborate with engineering teams to deliver performance improvements that benefit customers.
Requirements
- Deep understanding of operating system internals, including Linux kernel scheduler and memory subsystem.
- Proven experience identifying and resolving performance regressions in production environments.
- Hands-on expertise with profiling and tracing tools such as perf, eBPF, and VTune.
- Strong background in AI/ML workload performance and GPU/accelerator utilization.
- Experience with HPC workloads and job schedulers like Slurm or PBS.
- Familiarity with modern AI-first development workflows.
- Experience building automated performance testing and regression detection pipelines.
- Excellent analytical skills for designing experiments and drawing conclusions from data.
- Strong written and verbal communication skills for presenting findings.
- A collaborative mindset with the confidence to champion performance engineering.
Benefits
- Medical, dental, and vision insurance.
- Flexible paid time off.
- Employee stock options.
- Remote work; no travel required for most positions.