Embedded Software Vision Engineer

About Origin Origin (previously 10xConstruction) is building general-purpose autonomous robots for US construction to tackle rising costs, safety risks, and labour shortages. Our modular, multi-trade platform combines purpose-built hardware with real-time site intelligence to navigate complex environments and execute tasks with precision. Trained in high-fidelity simulation and already deployed on live sites, our robots deliver 5x faster execution, 250%+ margin expansion, and significant cost savings. Join India's most talent-dense robotics team consisting of individuals from IITs, Stanford, UCLA, etc.

About the Role You will own the electrical hardware that runs through the entire robot: custom PCBs and wire harnesses from power distribution through compute and networking down to microcontroller- based tool controllers. Your designs will span high-power boards (inverters, motor drives, power conversion), compute and camera interface boards with multi-sensor networking, and embedded microcontroller boards for end-effector tool systems. You will also own the wire harness architecture that ties it all together inside a mobile robot operating on dusty, vibration- heavy construction sites.

• Select and qualify camera hardware (sensors, lens assemblies, camera modules) against requirements from perception, navigation, manipulation, and controls teams ? making trade-off decisions on resolution, shutter type, dynamic range, FOV, focal length, spectral response, and latency characteristics for each use case on the robot.
• Bring up and maintain camera drivers on NVIDIA Jetson platforms (Orin, Orin Nano, Thor) for GMSL2, MIPI CSI-2, and GigE Vision (PoE) interfaces ? writing and patching V4L2 drivers, Linux kernel modules, device-tree overlays, and NVIDIA-specific components (libargus, nvarguscamerasrc, nv_multimedia_api).
• Develop and optimize real-time vision pipelines used in robot perception and control systems, ensuring deterministic performance and reliable operation under demanding field conditions.
• Debug and resolve imaging-system bottlenecks related to latency, jitter, frame drops, synchronization errors, memory bandwidth limitations, and frame-rate instability.
• Own optics and imaging quality ? specify lens parameters, tune ISP settings (white balance, exposure, gain, HDR), run intrinsic/extrinsic calibration for multi-camera rigs, and characterize camera performance under construction-site conditions including dust, vibration, spray mist, and highly variable lighting.
• Integrate and optimize depth and stereo vision camera systems, including calibration, synchronization, and performance characterization for robotics applications.
• Design and implement hardware triggering (PWM/GPIO) and PTP (IEEE 1588) synchronization to align frame capture across multiple cameras and onboard sensors, ensuring sub-millisecond cross-sensor timing accuracy.
• Build and optimize image-acquisition pipelines using GStreamer and NVIDIA DeepStream that move frames from sensor to GPU memory with zero-copy mechanisms (DMA-BUF, NVMM) and minimal processing overhead.
• Leverage NVIDIA GPUs and CUDA to accelerate image processing, vision algorithms, and AI inference workloads while maintaining real-time system constraints.
• Specify and integrate illumination systems (structured light, LED strobe) synchronized with camera exposure windows to improve image quality and sensing reliability in challenging environments.

• 3?5 years of hands-on experience with embedded camera systems, including sensor evaluation, camera driver bring-up, and imaging pipeline development on NVIDIA Jetson platforms (Xavier, Orin, or Orin Nano).
• Strong experience developing real-time vision-based systems for robotics, automation, machine vision, or autonomous platforms.
• Experience debugging system-level issues related to latency, jitter, throughput bottlenecks, frame synchronization, and frame-rate stability.
• Hands-on experience with depth cameras, stereo vision systems, or multi-camera imaging setups.
• Driver-level experience with GMSL2, MIPI CSI-2, or GigE Vision camera interfaces, including device-tree overlays, V4L2 subsystem, and low-level camera debugging.
• Experience developing or modifying Linux kernel drivers for camera interfacing on embedded Linux platforms and single-board computers.
• Working knowledge of optics and imaging fundamentals, including lens selection, exposure control, ISP tuning, distortion management, and camera calibration.
• Strong proficiency in C and C++ with the ability to read, debug, and modify Linux kernel and V4L2 code.
• Experience building low-latency image acquisition and processing pipelines using GStreamer, NVIDIA Multimedia API, DeepStream, or similar frameworks.
• Experience utilizing NVIDIA GPUs and CUDA for image processing, computer vision workloads, and model inference optimization.
• Strong understanding of Linux-based embedded systems and performance optimization techniques.

• Experience with industrial machine-vision cameras and the GigE Vision / GenICam ecosystem (FRAMOS, Basler, FLIR, Lucid).
• Experience with active illumination systems such as structured light, synchronized LED strobes, or other machine-vision lighting solutions.
• Hardware triggering and multi-sensor synchronization using PTP (IEEE 1588), GPIO, or PWM-based trigger architectures.
• Familiarity with ROS 2 camera drivers, robotics middleware integration, and deployment of camera systems on autonomous robots.
• Experience optimizing end-to-end camera-to-GPU pipelines using zero-copy architectures and CUDA-accelerated processing workflows.