SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Path Robotics is seeking an experienced Perception Engineer to own the end-to-end real-time perception pipeline for intelligent robotics systems. This role focuses on building robust vision systems that close control loops on embedded hardware, addressing one of the most challenging aspects of autonomous robotics: reliable perception in degraded or adversarial conditions.
Key responsibilities include:
- Designing and implementing the complete real-time perception pipeline from sensor input through feature extraction and geometric estimation to control loop output, operating at specified frame rates on embedded hardware within strict computational budgets.
- Developing robust perception algorithms that maintain accuracy when sensor data is deliberately or naturally degraded, and implementing confidence estimation so the system knows when estimates are unreliable.
- Performing geometric estimation against known models to recover object positions and orientations from partial and imperfect sensor data, with proper error propagation and failure detection.
- Evaluating and selecting sensing hardware options through rigorous experimentation, defending choices with quantitative measurements rather than specifications.
- Setting technical direction as the perception work evolves toward learned methods and more sophisticated approaches.
Required qualifications:
- Proven experience closing real-time perception loops on physical hardware with cameras, including demonstrated knowledge of end-to-end latency, frame rates, and measured system performance.
- Deep expertise in classical computer vision and 3D geometry: camera models, triangulation, registration, calibration, and proper error propagation.
- Production experience with real-time segmentation or feature extraction on degraded imagery, deployed on real systems, with understanding of why standard metrics fail on thin or low-contrast structures.
- Strong modern C++ and ROS 2 proficiency.
- Experience with GPU-accelerated inference on embedded platforms under hard per-frame budgets, with attention to tail latency rather than average performance.
- Ability to defend accuracy claims with independent ground truth measurements, reporting bias and repeatability separately.
This is a hands-on technical role for someone who enjoys solving hard problems at the intersection of classical and learned perception methods.
