Senior/Staff System Software Engineer - ADAS Vision Platform

• Own the runtime infrastructure of a real-time vision system: task orchestration, inter-process messaging, sensor ingestion, and deterministic data flow under hard latency budgets
• Build the infrastructure that lets the vision stack run anywhere - on target ECUs, on a developer’s desk, and at scale against large volumes of recorded drive data
• Design and evolve data replay, simulation, and regression-testing systems that turn field recordings into fast, repeatable development loops
• Keep the stack portable and efficient across compute platforms and DNN inference runtimes, from datacenter GPUs to automotive SoCs
• Design the real-time communication paths that connect the vision system to vehicle platforms and external consumers
• Engineer to automotive safety and security standards (ISO 26262, ISO 21434) without giving up development speed
• Build and sharpen the team’s AI-assisted engineering workflows - coding agents, AI-assisted code review, test generation - and set the standard for how they’re used well

• 6+ years building system software for real-time or embedded environments
• Expert-level C++ (and solid C), including debugging on real hardware under real-time constraints
• Deep grounding in operating systems and computer architecture: scheduling, memory, caches, IPC, and what they cost
• Hands-on ROS or ROS2 experience - designing nodes/components, defining messages, working with bag data
• Fluency in serialization and middleware (Protocol Buffers, DDS, ZeroMQ, or similar) and network programming over TCP/UDP/Ethernet
• Working practice with AI-assisted development tools as a normal part of how you build, debug, and test
• BS/MS in Computer Science or equivalent experience

Nice to have

• Depth in an RTOS or safety-grade platform: QNX, RT Linux, or Green Hills INTEGRITY
• Experience across both ROS1 and ROS2 ecosystems, including bag formats (rosbag/rosbag2/MCAP) and DDS configuration
• DNN inference runtimes and accelerators: TensorRT, TIDL, CUDA, quantized inference, and their pre/post-processing pipelines
• Cloud-scale data or simulation pipelines: containers, orchestration, batch compute, CI at fleet-data volume
• Automotive communication protocols: CAN, automotive Ethernet, SOME/IP
• Time spent shipping software in an automotive, robotics, or autonomous-vehicle setting

PhantomVision is Phantom AI’s production computer vision system for ADAS: a multi-camera, real-time perception stack that ships on automotive ECUs running Linux, QNX, and INTEGRITY. Behind the perception algorithms sits the platform that makes them run - the real-time runtime, the sensor and data pipelines, the simulation and test infrastructure, and the interfaces to the vehicle. This role owns that platform.