Software Engineer - Realtime Pipeline

Bridge Neurotech is seeking a software engineer to own a tightly coupled, end-to-end realtime pipeline for neural decoding.

Responsibilities

• Own the full realtime path: raw RF data ingestion, beamforming, processing (including clutter filtering), ML preprocessing, model inference, and decoded output
• Build and mature the ML inference engine, including:
  • Model loading from artifacts
  • GPU memory residency management
  • Batching strategy for latency-constrained inference
  • TensorRT/ONNX runtime integration or hand-built CUDA inference
  • Kernel-level integration with upstream imaging compute
• Integrate, extend, and contribute to GPU beamforming where appropriate
• Meet and maintain realtime throughput and latency goals at multi-Gbit/s ingest with deterministic tail latency suitable for closed-loop decoding
• Define and own the model serving contract with the ML/AI Lead, including a path for model updates without interrupting service
• Build the streaming and storage layer to support raw and processed data for both offline training and realtime inference
• Collaborate to define upstream interfaces to the device acquisition layer and downstream interfaces to BCI applications
• Implement observability across every stage and create benchmarks and regression tests to lock in pipeline performance and inference correctness as the system evolves

Requirements

• BS/MS/PhD in computer science, electrical engineering, or related field
• 8+ years building production realtime or HPC systems
• Expert-level CUDA and modern C++, including experience writing or extensively modifying GPU kernels for signal processing or imaging workloads
• Proven experience building production ML inference systems by deploying trained PyTorch (or equivalent) models to GPUs under hard latency budgets, including hands-on work with TensorRT, ONNX Runtime, Triton, or hand-built CUDA inference engines
• Demonstrated realtime data pipeline experience at multi-Gbit/s rates with tight latency budgets
• Strong grasp of data structures, algorithms, and signal processing for imaging or RF, including:
  • Beamforming concepts
  • FFTs
  • IQ demodulation
  • SVD-based filtering
  • Autocorrelation
• Proficiency in Python and the PyTorch ecosystem to collaborate with the ML team, identify production-impacting changes, and discuss model architecture trade-offs that affect latency

Technologies

• CUDA, C++, PyTorch, TensorRT, ONNX, ONNX Runtime, Triton
• Python, FFT, IQ demodulation, SVD-based filtering, autocorrelation

Location

• Redwood City, CA (onsite)

Benefits

• Excellent medical benefits
• 401(k) with 6% match

Nice to Have

• Experience with functional imaging or other high-data-rate imaging modalities
• Experience integrating ML inference with upstream realtime compute, including shared GPU memory layouts and minimizing host-device round-trips
• Experience with high-bandwidth I/O: GPUDirect Storage, RDMA, kernel-bypass networking, NVMe arrays