| Company | Global quantitative trading firm |
| Company Info | A global quantitative trading firm founded with an entrepreneurial mindset and a rigorous analytical approach to decision making. |
| Job Title | GPU Kernel Performance Engineer |
| Job Short Desc | Build ultra-low-latency GPU inference systems, designing custom CUDA kernels and execution strategies for sophisticated quantitative workloads where deep hardware knowledge and creative performance engineering create measurable impact.
|
| Salary | Variable based on experience |
| Locations | APAC - Hong Kong (China) |
| Visa | Y |
| Job Summary | Join a highly technical GPU performance team focused on building ultra-low-latency inference systems for sophisticated quantitative workloads. You’ll design and optimize custom CUDA kernels, fine-tune memory layouts and execution strategies, and profile GPU performance to uncover opportunities beyond standard runtimes and vendor libraries. Working closely with quantitative researchers and engineers, you’ll translate mathematical models into production-grade GPU implementations across neural networks, tree-based models, state space models, and other structured workloads. The role combines deep GPU architecture knowledge with numerical reasoning, low-level systems programming, and creative performance optimization, offering the opportunity to influence both model design and the hardware-efficient systems that execute it.
|
| Criteria | | Engineer custom GPU solutions for latency-critical inference. | | Optimize kernels, memory layouts, and execution strategies. | | Partner with researchers to accelerate complex models. |
|
| Apply | Apply |