Visual SLAM Researcher

DeepReach AI; San Francisco (CA); On-site; USD 150,000 - 210,000; Full time

Job summary

DeepReach AI, an early-stage physical AI company, builds data infrastructure for robot learning and collaborates with frontier AI labs, robotics firms, and leading universities. The role focuses on taking SLAM/VIO/SfM innovations from research to robust production pipelines across devices, with calibration, synchronization, and validation at the core.

You’ll partner with hardware teams to test on-site and push reliable spatial computing into real products.

Qualifications

• 3+ years of experience in SLAM or 3D vision (graduate research counts).
• Strong foundation in multi-view geometry and state estimation.
• Proven experience building, maintaining, or improving SLAM, VIO, and SfM systems.
• Proficiency in C++ and Python; familiarity with OpenCV, Ceres, GTSAM, COLMAP, Kalibr or similar.
• Highly hands-on and product-oriented: you turn algorithms into reliable production systems.

Responsibilities

• Own, maintain, and improve multiple SLAM/VIO/SfM pipelines across capture devices.
• Run rigorous multi-camera and IMU calibration and time synchronization.
• Turn algorithms into high-throughput offline pipelines with quality metrics.
• Build visualization and debugging tools for images, point clouds, and trajectories.
• Serve as a computer vision hub for pre-/post-processing, data validation, and labeling.
• Work closely with the hardware team on sensor calibration and on-site testing.

Skills

Multi-view geometry

State estimation

Hands-on debugging

Production systems

English communication

Tools

C++

Python

OpenCV

Ceres

GTSAM

COLMAP

Kalibr

PyTorch

FFmpeg

Rerun

Job description

We are an early-stage physical AI company building the data infrastructure for robot learning. We work closely with frontier AI labs, robotics companies, and leading university research groups.

What You'll Do

• Own, maintain, and improve multiple visual-inertial SLAM / VIO / SfM pipelines across capture devices (custom hardware, phones, and more).
• Run rigorous multi-camera and IMU calibration and time synchronization to keep spatial computing consistent and robust across devices.
• Turn algorithms into high-throughput offline pipelines at scale, with trajectory quality metrics and automatic failure detection.
• Build visualization and debugging tools for images, point clouds, and trajectories.
• Serve as a computer vision hub for the team, supporting pre- and post-processing, data validation, and automated labeling.
• Work closely with the hardware team on sensor calibration and on-site testing.

What We're Looking For

• 3+ years of experience in SLAM or 3D vision (graduate research counts).
• Strong foundation in multi-view geometry and state estimation: camera models, bundle adjustment, factor graphs, filtering.
• Proven experience building, maintaining, or improving SLAM, VIO, and SfM systems.
• Proficiency in C++ and Python; familiarity with OpenCV, Ceres, GTSAM, COLMAP, Kalibr or similar.
• Highly hands-on and product-oriented: you turn algorithms into reliable production systems.

Nice to Have

• Calibration and SLAM with fisheye or ultra-wide cameras.
• Processing mobile SLAM / AR data (e.g., ARKit, ARCore).
• Learning-based SLAM, feed-forward 3D reconstruction, or neural reconstruction (3DGS / NeRF).
• Optical flow or motion modeling for head-mounted cameras under motion blur and fast rotation.
• Familiarity with PyTorch, FFmpeg, Rerun or similar tools.
• Publications at CVPR, ICCV, ECCV, ICRA, IROS or similar.

What Success Looks Like

• SLAM / SfM systems run reliably and efficiently on every capture device.
• Customers trust our spatial data because the underlying algorithms and tooling are rock solid.

Why Join Us

• Work alongside frontier AI labs and top university research groups on the field's hardest problems.
• Publish at top venues and contribute to open source, with full support.
• Own your work end to end, from sensors to algorithms to final delivery.
• Early team, high ownership, competitive salary and equity.