GPU Performance Engineer
Core
Design, implement, and optimize custom CUDA kernels for low-latency inference of structured models (neural networks, tree-based, state space) in production systems.
Role type
Senior IC GPU Performance Engineer
Builds
Production-grade GPU inference pipelines for quantitative trading workloads
Domain
Financial technology / High-performance computing
Deliverable
production ML models
Required skills
CUDA kernel optimization, C/C++, GPU architecture (memory hierarchy, SIMT), low-level systems programming, performance profiling, numerical stability analysis
Preferred skills
PTX-level tuning, tensor core utilization, model compilation, kernel fusion, experience with ONNX Runtime/TensorRT/Triton/TVM, linear algebra, quantitative research background
Technologies
CUDA, C/C++, PTX, TensorRT, Triton, TVM, ONNX Runtime, KightGBM, Mamba architectures
Responsibilities
Design and optimize custom CUDA kernels for latency-critical inference; Analyze computational bottlenecks in quantitative research models; Collaborate with researchers to translate mathematical models into high-performance compute pipelines; Profile and benchmark GPU performance; Contribute to GPU architecture decisions and performance best practices
Seniority
Senior, hands-on IC
