About the Role
- Design, build, and own an end-to-end high-throughput, low-latency production system.
- Build real-time video pipelines that include streaming, processing, and advanced CV inference.
- Optimize GPU inference performance (latency/throughput/memory) and overall resource usage
- Deep-dive debugging in production environments to resolve performance bottlenecks
- Build for reliability: logging, monitoring, graceful degradation, and fault tolerance
- Design and implement the deployment flow: Docker packaging, installation, and orchestration (K8s or similar).
Requirements
Strong software engineering background with a focus on production systems (Python and C++ a plus).
Experience developing and maintaining production services, including debugging, monitoring, and ensuring reliability.
Performance-oriented approach: profiling, identifying bottlenecks, and managing memory and resources.
Proven experience with containerization (Docker), system orchestration, and cloud platforms like GCP or AWS.
Nice to Have:
Experience with GPU inference frameworks such as PyTorch, TensorRT, and ONNX Runtime is a major advantage.
A strong background in video streaming technologies like RTSP, GStreamer, and FFmpeg is highly desirable.
Familiarity with the NVIDIA stack, including DeepStream, TensorRT, CUDA profiling tools, and Jetson/Orin platforms, is preferred.
Kubernetes expertise is also important.
Proven experience collaborating closely with research or data science teams in a deep-tech startup environment is valued.
About the Company
MotionAnalytics develops MotionID, a breakthrough biomechanical AI that identifies people by how they move, not by their faces, clothing, or devices.
Based on over two decades of biomechanics research at Ben-Gurion University, MotionID transforms standard video into unique biomechanical signatures that enable reliable, hardware-free identification of the same individual across different videos. The technology performs in real-world conditions where traditional biometrics struggle, including when faces are covered, lighting is poor, and cameras are thermal, distant, ground-based, or aerial.
Join us in developing the world’s first large biomechanical foundation model. We are looking for a hands-on engineer skilled in designing mission-critical, engineering-grade systems that enable real-time operation of computer vision models.