Software Engineer – Physical AI Platform Integration

About the role

• We are looking for a hands-on software engineer to make Lyte's LiDAR and camera perception systems work end to end on real robots and on the compute platforms our customers use. You will own the software path from sensor packets on the wire to calibrated, time-synchronized point clouds and images inside a robotics stack: sensor drivers and SDK, GPU-accelerated processing, calibration and end-of-line test tooling, data recording and playback, and the containers and CI/CD that let a small team ship all of this quickly and reliably. Much of the work is turning early results into maintained, reusable components that robotics teams inside and outside Lyte can depend on: a proof of concept becomes a driver, a one-off script becomes a calibration pipeline, a demo becomes a supported release. You will work closely with algorithm, hardware, silicon, and manufacturing engineers, and split your time between building new capability and supporting the people who use it. Expert-level C++ and Python are required. You should be equally at home in a CUDA kernel, a ROS 2 node, a CMake file, and a GitLab pipeline.

What you'll do

• Develop and maintain sensor drivers, ROS 2 packages, and SDK components that deliver Lyte LiDAR and camera data (point clouds, Doppler velocity, images, IMU) to robotics and perception stacks on NVIDIA Jetson and x86 GPU platforms.
• Build GPU-accelerated processing on the receiving side: point-cloud filtering and clustering, camera-to-LiDAR projection and image rectification, motion compensation, and ego-velocity or odometry estimation, packaged as reusable C++ libraries with Python bindings.
• Own time synchronization and data-path correctness for Ethernet-attached sensors: PTP, packet parsing, times-tamping, and frame assembly for high-bandwidth streams (for example NVIDIA Holoscan Sensor Bridge, RoCE/UDP).
• Design and maintain calibration, KPI, and end-of-line test pipelines used in manufacturing and field bring-up, including automation of test fixtures such as robotic arms and data-acquisition stations.
• Build the recording, playback, and visualization tooling (pcap, MCAP, Foxglove, compressed video) that engineers and customers use to capture, replay, and debug sensor data.
• Own the developer platform: Docker images for multiple CPU/GPU targets, ROS 2 distributions, Python and C++ packaging (wheels, .deb), and CI/CD on GitLab CI and Jenkins (Groovy pipelines and shared libraries) with fast, risk-based gating, hardware-in-the-loop test stages, and release automation.
• Act as first responder for integration issues raised by internal teams, technology demonstrations, and early customers; reproduce problems from logs and recordings, fix root causes, and feed the lessons back into tooling and documentation.
• Work in tight loops with hardware, silicon, algorithm, and manufacturing teams across time zones; write clear documentation and modular code that others can extend.
• Use AI-assisted coding tools to move faster without sacrificing review quality or correctness.

Required Qualifications

• Bachelor's degree in Computer Science, Electrical Engineering, Robotics, or a related technical field, or equivalent practical experience.
• 3+ years of professional software engineering experience on systems that interface with real sensors or hardware.
• Expert-level modern C++ (C++17 or newer) and Python, including CMake-based builds and Python packaging.
• Hands-on experience with ROS 2 (or comparable robotics middleware) and Linux systems programming, including networking and multi-threaded, real-time data paths.
• Working knowledge of CUDA or another GPU compute framework, and of deploying software on NVIDIA Jetson or similar embedded GPU platforms.
• Proven ability to build and maintain Docker-based development environments and CI/CD pipelines, in GitLab CI and Jenkins (declarative pipelines and shared libraries in Groovy) or equivalent.
• Track record of taking prototypes to maintained, tested, documented components; comfortable owning a component end to end in a fast-moving, sometimes ambiguous environment.
• Strong communication and collaboration skills across disciplines and time zones.

Preferred Qualifications

• Experience integrating LiDAR or camera sensors: point-cloud processing, camera models and rectification, intrinsic and extrinsic calibration, sensor time synchronization (PTP, hardware triggers).
• Experience with NVIDIA Holoscan, Isaac ROS / NITROS, TensorRT, or similar accelerated perception frameworks.
• Background in state estimation or odometry (ego-motion, visual-inertial odometry, factor-graph optimization) or in signal processing for FMCW / Doppler sensing.
• Experience with robotics data tooling: MCAP/rosbag, Foxglove, pcap analysis, large-dataset ingestion (Parquet).
• Familiarity with manufacturing or end-of-line test environments, robotic test fixtures, and release and versioning discipline for production images.
• Experience with Python/C++ interop (nanobind, pybind11, Cython), profiling (Tracy, Nsight), and micro-benchmarking.
• Experience running hardware-in-the-loop CI: Jenkins agents attached to device benches, Groovy shared libraries that drive flashing, power control, and on-device tests, and Gerrit or GitLab triggered device pipelines.
• Startup or small-team experience; comfort supporting demos and customers directly.
• MS or PhD in a related field is a plus, not a requirement.

Benefits (subject to location and local regulations)

• Competitive salary and equity
• Comprehensive medical, dental, and vision coverage
• 401(k) retirement plan
• Flexible vacation and time-off policy
• Collaborative, fast-paced, and inclusive work environment
• Opportunity to work on cutting-edge technologies with a highly cross-functional team