Realtime Neural Pipeline Engineer

What you'll do

• Build and optimize the realtime path, including RF ingestion, beamforming, processing, ML preprocessing, inference, and decoded output.
• Set technical direction; develop GPU inference and beamforming, streaming and storage, and model-serving interfaces.
• Meet latency and throughput targets, and build observability, benchmarks, and regression tests.

What they ask for

• Production ownership of realtime or HPC systems and experience building multi-Gbit/s realtime data pipelines under tight latency budgets.
• Expert-level CUDA and modern C++, including writing or extensively modifying GPU kernels for signal processing or imaging.
• Production GPU ML inference experience, deploying trained models under hard latency budgets; hands-on with TensorRT, ONNX Runtime, Triton, or custom CUDA inference.
• BS/MS/PhD in computer science, electrical engineering, or a related field; 8+ years of work experience; proficiency in Python and PyTorch, plus relevant imaging or RF signal-processing knowledge.

Nice to have

• Experience with functional ultrasound or other high-data-rate imaging modalities.
• Experience integrating ML inference with realtime compute or with high-bandwidth I/O.

Tools

CUDA, C++, PyTorch, TensorRT, ONNX Runtime, Triton, GPUDirect Storage, RDMA

AI focus

Deploy neural decoding models for realtime BCI outputs, including GPU inference, model serving, and coordination with the ML/AI team on latency-affecting architecture trade-offs.