Staff Computer Vision Engineer

Point One Navigation Inc•San Francisco, CA

7d•$215,270 - $265,800

About The Position

Staff Computer Vision Engineers are responsible for the comprehensive lifecycle of Point One’s spatial AI and visual navigation features, overseeing everything from initial camera integration and image processing to high-level architectural and algorithmic design. This is an ownership-first role: you will conceptualize and drive complex technical challenges end-to-end - from early architecture through deployment in mission-critical systems - while raising the technical bar across the team. FusionEngine already powers a wide range of devices, hardware platforms, and customer applications. The R&D team is responsible for making sure our vision and perception systems work reliably across all of them: Designing solutions robust to visually challenging environments, optimizing models for compute-constrained edge devices, and ensuring our algorithms stay thoroughly tested, verified, and production-ready as we scale. Success in this role means: State-of-the-art CV and SLAM techniques are successfully translated from research papers, internal prototypes, or third-party solutions into highly performant, production-grade algorithms. Rigorous benchmarking pipelines are established to objectively evaluate internal algorithms against commercial OTS solutions and vendor offerings. Vision pipelines automatically generate and maintain accurate, semantically rich maps of complex indoor environments with minimal manual intervention. Real-time localization and multi-agent tracking (assets, robots, people) are highly robust, minimizing latency and identity switches even in dynamic or visually degraded conditions. Spatial data, coordinate frames, and map layers are exposed via clean data models and APIs, empowering our UI and infrastructure teams to build seamless user-facing applications. Junior engineers grow faster and the team's practices improve measurably over time. Immediate Areas of Focus Applied Research, Benchmarking & Selection Lead the research, evaluation, and selection of state-of-the-art computer vision, deep learning, and spatial navigation methodologies for highly accurate 3D maps of large-scale facilities, considering both internal development and third-party commercial solutions. Develop or integrate deep learning and classical CV algorithms to extract semantic information from environments (e.g., structural elements, zones, and specific objects) for overlay onto base map. Ensure maps can be dynamically updated over time as the physical layout of a facility changes, enabling map version management and consistency. Design and own a rigorous benchmarking framework to continuously evaluate the accuracy, latency, compute footprint, and reliability of internal code versus off-the-shelf and vendor technologies. Rapidly prototype new perception capabilities and architect their transition into highly optimized, edge-capable production code, or seamlessly encapsulate and integrate verified third-party modules. Collaborate tightly with infrastructure and UI engineers to manage data products, render maps, and track assets for the end user. Drive Real-Time Localization and Tracking Understand how and work with the larger navigation team to use camera data with GNSS, IMU, wheel odometry, and other indoor positioning signals to maintain high-confidence state estimation for moving agents in all environments. Drive performance tuning for edge deployment to ensure tracking algorithms run with low latency and high reliability on constrained compute architectures. Proactively identify failure modes in tracking and mapping and design robust algorithmic fallbacks. Raise the Technical Bar Mentor junior engineers and establish best practices across the team. Contribute to architecture discussions, technical strategy, and roadmap planning.

Requirements

7+ years of professional algorithm and software development experience, with significant depth in applied research, computer vision, or robotics.
Expertise in modern C++ (C++14 or later) and Python.
Demonstrated history of success of taking AI model prototypes (PyTorch, TensorFlow) and turning them into scalable, real-time production systems.
Expertise in ROS1/ROS2.
Hands-on experience with Visual SLAM, 3D reconstruction, and mapping architectures.
Experience in deploying semantic segmentation/object detection in real-world environments.
Experience with multi-view geometry, camera calibration, and fusing vision with other sensor modalities (IMU, GNSS).
Ability to take high-level research and business goals and decompose them into actionable engineering tasks, realistic schedules, and clear milestones.
MS or PhD in Computer Science, Robotics, or equivalent experience.

Nice To Haves

Background in deploying optimized vision models to edge devices using TensorRT, ONNX, or platform-specific accelerators.
Experience in deploying multi-object tracking and ReID architectures in real-world, dynamic environments.
Familiarity with managing large-scale point clouds, mesh generation, or NeRFs/Gaussian Splatting for environmental representation.

Responsibilities

Oversee the comprehensive lifecycle of spatial AI and visual navigation features, from initial camera integration and image processing to high-level architectural and algorithmic design.
Conceptualize and drive complex technical challenges end-to-end, from early architecture through deployment in mission-critical systems.
Raise the technical bar across the team.
Design solutions robust to visually challenging environments.
Optimize models for compute-constrained edge devices.
Ensure algorithms are thoroughly tested, verified, and production-ready.
Translate state-of-the-art CV and SLAM techniques into performant, production-grade algorithms.
Establish rigorous benchmarking pipelines to evaluate internal algorithms against commercial solutions.
Develop vision pipelines that automatically generate and maintain accurate, semantically rich maps of complex indoor environments.
Ensure real-time localization and multi-agent tracking are highly robust, minimizing latency and identity switches.
Expose spatial data, coordinate frames, and map layers via clean data models and APIs.
Mentor junior engineers and establish best practices.
Contribute to architecture discussions, technical strategy, and roadmap planning.
Lead the research, evaluation, and selection of computer vision, deep learning, and spatial navigation methodologies.
Develop or integrate deep learning and classical CV algorithms to extract semantic information from environments.
Ensure maps can be dynamically updated over time.
Design and own a rigorous benchmarking framework.
Rapidly prototype new perception capabilities and architect their transition into production code.
Collaborate with infrastructure and UI engineers to manage data products, render maps, and track assets.
Use camera data with GNSS, IMU, wheel odometry, and other indoor positioning signals to maintain high-confidence state estimation.
Drive performance tuning for edge deployment.
Proactively identify failure modes in tracking and mapping and design robust algorithmic fallbacks.