Senior Perception Engineer | Sensors, Firmware & Drivers

Job description

Job responsibilities

• Bring up new imaging and depth sensors end to end: I2C/SPI enumeration, register maps, GMSL2/GMSL3 and FPD-Link serializer/deserializer configuration, CSI-2 lane mapping, and first light on target hardware.
• Write and maintain V4L2 / MIPI CSI-2 camera drivers and device tree overlays for NVIDIA Jetson platforms (Orin, Thor), including kernel work against the Jetson Linux BSP.
• Build low-latency, zero-copy sensor ingest and pre-processing pipelines in NVIDIA Holoscan, GStreamer, and CUDA | ISP tuning, debayer, rectification, format conversion with a hard budget on glass-to-algorithm latency.
• Own multi-sensor time synchronization: PTP/gPTP, hardware trigger distribution, exposure-to-timestamp accounting, and drift characterization across cameras, lidar, IMU, and robot joint state.
• Partner with EE on schematic and layout review for sensor boards, GMSL coax/harness design, power sequencing, and connector selection; debug with a scope, logic analyzer, and thermal camera in hand.
• Characterize and qualify sensors for the real environment: welding-arc dynamic range, IR interference, vibration, ingress, temperature soak, and EMI from high-current equipment.
• Build the calibration and validation infrastructure | intrinsics/extrinsics, HDR and exposure behavior, automated bench tests, CI on real hardware.
• Work directly with vendors and silicon partners on errata, firmware revisions, and early access to new parts.

Job requirements

• 5+ years building embedded sensor, firmware, or driver systems in robotics, autonomous vehicles, aerospace, or comparable hardware-heavy domains.
• Strong C/C++ and Python; comfortable in the Linux kernel.
• Bonus: Hands-on experience with NVIDIA Jetson platforms, Jetson Linux BSP, device tree, kernel module development, flashing and board support work.
• Demonstrated camera bring-up experience: MIPI CSI-2, V4L2, GMSL2/GMSL3 (Maxim/ADI) or FPD-Link, I2C/SPI sensor configuration.
• Experience shipping software that runs on hardware in the field, with the instincts about failure modes that comes from having done it.
• NVIDIA Holoscan SDK, DeepStream, or CUDA/GPU-accelerated streaming pipelines.
• Time synchronization across heterogeneous sensors (PTP, hardware trigger design).
• ISP tuning, HDR imaging, or low-level camera sensor characterization.