Source description
About the role
About Neurabit Neurabit is a bootstrapped, DPIIT-recognised, deep-tech company building edge AI, computer vision, and physical AI systems that run where the internet doesn't. Our systems are deployed across 7+ countries from passenger analytics on public transit fleets to industrial safety monitoring, substation surveillance, and sports analytics. We're a team of ~18 engineers who ship real hardware-software systems into messy, real-world environments: moving buses, dusty factory floors, remote substations. We're not a research lab. Models here don't live in notebooks they live on Jetsons bolted inside vehicles and on factory walls, running 24/7. The Role You'll own computer vision systems end to end: from architecture and model selection through optimization, edge deployment, and production hardening. You'll work directly with the founding team on flagship projects multi-camera passenger counting, PPE and zone compliance detection, ANPR, crowd intelligence, and driver monitoring and your decisions will ship to paying customers within weeks, not quarters. What You'll Do - Design and ship multi-camera CV pipelines for detection, tracking, and counting in constrained edge environments - Build and tune object tracking systems (ByteTrack, DeepSORT, or similar) for occlusion-heavy, real-world scenes - Optimize models for edge inference quantization, pruning, TensorRT/ONNX conversion targeting NVIDIA Jetson (Orin/Xavier/Nano), Hailo, and similar accelerators - Own accuracy in the field: design evaluation protocols, debug failure modes from real deployment footage, and close the gap between lab metrics and production performance - Work with RTSP/RTMP camera streams, handle unreliable networks, and build pipelines that degrade gracefully - Write clean, maintainable Python/C++ and mentor mid-level engineers on CV best practices - Collaborate with hardware, backend, and product teams to scope what's technically feasible before we commit it to a client Must-H .