Lead Robotics Software Engineer, Localization

Rethink RecruitSan Francisco, CaliforniaOn-siteFull-timeSenior, 5–8 yearsListed 1 hour ago

Apply now

About this role

ABOUT THESEUS

Theseus is building a Visual Position System that is 10x lighter and cheaper than the nearest competitor. Drones need to know where they are, but GPS is a luxury on the modern battlefield. Theseus built Cyclops, their visual localization software, to enable precise, autonomous navigation without GPS.

Just two years in, Theseus has customers across every branch of the U.S. military and major defense primes, with the product actively used in combat in Ukraine. The company expects to sell more than 10,000 licenses this year. The next challenge: take that capability across ground, air, and maritime environments to build a universal alternative to GPS.

THE OPPORTUNITY

Theseus is looking for robotics engineers with SLAM, VIO, localization, or state estimation experience to work on the estimator itself: the visual-inertial odometry, the map matching, and the sensor fusion that turn camera frames and IMU samples into a position the flight controller can trust.

Every day, you will work on feedback beamed in from real deployments across the world, and your changes will ship to devices in the field every single week. You will lead much of the localization work at Theseus.

WHAT YOULL DO

- Work on the visual-inertial odometry, the map matcher, and the filter that fuses them, so that Cyclops holds an accurate position in daylight, at night, over featureless terrain, and on IR cameras

- Build features that increase availability and reduce operator setup, including faster initialization, recovery after lost estimates, better handling of poor calibration, and honest uncertainty reporting

- Take problems that arrive as flight logs from customer deployments, reproduce them in replay, form a hypothesis about why the estimate diverged, fix it, and prove the fix across the recorded flight library before it ships

- Prototype next-generation algorithms aimed at order-of-magnitude gains in accuracy, robustness, and availability, tested against real data

- Ship regularly at a weekly cadence: new features and fixes go out fast, and new models ship within days to weeks

YOU SHOULD HAVE

- 3+ years (mid-level) or 5+ years (senior) building and deploying state estimation, SLAM, or visual-inertial odometry systems that shipped in a product or were validated in field testing

- Depth in estimation fundamentals: you can derive an EKF or a factor graph on a whiteboard, reason about frames and rotations on SO(3) and SE(3), explain IMU preintegration, and discuss observability, consistency, and covariance behavior when a sensor drops out

- A working grasp of the geometry underneath visual localization: camera models including fisheye and thermal, calibration, epipolar geometry, feature tracking, and bundle adjustment

- Experience with at least one modern estimation toolchain such as GTSAM, Ceres, OpenVINS, ORB-SLAM, or VINS, with enough internal understanding to modify rather than just configure it

- Production C++ in a Linux environment, with comfort reasoning about latency, memory, and compute cost on ARM edge hardware

- An engineering mindset focused on outcomes: you know the difference between the fix that cleans up this week's logs and the redesign that makes next year's product better

NICE TO HAVE

- Experience with camera and IMU calibration

- Map matching, bundle adjustment, or georegistration against aerial or satellite imagery

- Experience with drones, ArduPilot, or MAVLink

- Real-time deployment on embedded devices

COMPENSATION AND BENEFITS

Base Salary: Up to $300,000, depending on experience

Theseus offers equity compensation along with medical, dental, and vision insurance, paid time off, and additional benefits.