Role Overview
As part of Tower Research's Core Engineering team, you will bridge quantitative research and high‑performance production systems, architecting inference pipelines that operate at the physical limits of hardware to achieve microsecond‑level latency.
Responsibilities
- Lead technical evaluation of CPU, GPU, and FPGA inference platforms to guide deployment decisions.
- Analyze and enhance memory hierarchy execution to maximize utilization and resolve bottlenecks across the inference lifecycle.
- Collaborate with infrastructure teams to design latency‑critical inference strategies within thermal, power, and operational constraints.
- Develop optimized GPU kernels and integrate performance libraries to achieve maximum computational throughput.
- Apply model reduction techniques (quantization, pruning, distillation) for low‑latency, event‑level inference.
- Work with ML researchers, HPC, FPGA, and datacenter engineers to deliver target deployments.
Requirements
- 2+ years experience optimizing deep learning inference in latency‑sensitive or high‑throughput production environments.
- Deep expertise in lower‑level ML framework development (PyTorch/JAX) with strong Python/C++ and mixed‑precision computation knowledge.
- Proven experience in custom GPU kernel development and familiarity with optimization libraries (Triton, TensorRT, ONNX, IREE, HLS4ML, cuBLAS, CUTLASS) and profiling tools (Nsight Systems, Nsight Compute).
- Deep expertise in GPU microarchitecture, including SM execution, warp scheduling, and full memory hierarchy optimization.
- Proven record of rigorous, data‑driven benchmarking across heterogeneous compute architectures.
- Bonus: practical experience optimizing inference on specialized hardware such as FPGAs and ASICs.
- Financial trading experience is not required.