PNT Software Engineer

Ulysses is the ocean company. We design, manufacture, and operate autonomous vehicles for surface and subsea operations — taking on the most critical tasks across defense, commercial, and scientific missions. Our vertically integrated platform includes Mako, a modular AUV with 72-hour endurance and more onboard compute than any vehicle in its class; Leviathan, an autonomous surface vessel mothership; and Kraken, our automated launch-and-recovery system that closes the loop between surface and subsea with zero human intervention.

The ocean covers 71% of the Earth's surface and drives a $2.6 trillion global economy, yet it remains the least monitored domain on the planet. We're changing that — building autonomous systems that collapse the cost of ocean operations by orders of magnitude so that persistent coverage becomes the baseline, not the exception. We operate globally today across defense, ocean science, and commercial survey, working with partners including the U.S. Navy, the Government of Australia, the Great Barrier Reef Foundation, and The Nature Conservancy.

Founded in 2023 and headquartered in San Francisco, Ulysses has raised $48 million in funding from Andreessen Horowitz, Booz Allen Ventures, Harpoon Ventures, Pebblebed, Lowercarbon Capital, and others. Our team is lean, hands-on, and fast — we prototype in the shop, test on the water, and ship. If you want to work at the frontier, there's no frontier bigger than the ocean.

ABOUT THE ROLE

We're hiring a PNT (Positioning, Navigation, and Timing) Software Engineer to give our vehicles an accurate, reliable understanding of where they are and what's around them — above and below the waterline.

This role will primarily focus on localization technologies, including inertial navigation, DVL, and GNSS, while contributing to our broader vision and acoustic development efforts. You'll work closely with our electrical, robotics, and mechanical engineering teams across Mako, Leviathan, and Kraken.

WHAT YOU'LL DO

• Navigation and state estimation: architect and implement tightly-coupled navigation for subsea and surface vehicles using factor graphs. Fuse IMU, DVL, pressure, USBL/SBL acoustic positioning, and GNSS/RTK when surfaced, and keep position error bounded across multi-day dives.

• Docking and recovery perception: contribute to vision and acoustic localization for our most critical moments — Mako finding and docking with Kraken for fully autonomous launch and recovery, and Leviathan operating safely around other vessels.

• Surface situational awareness: fuse radar, AIS, and EO/IR cameras into a reliable track picture that drives collision avoidance and COLREGs-compliant autonomy on Leviathan.

• Calibration: own IMU, DVL, and GNSS calibration end-to-end — IMU bias, scale-factor, and misalignment, DVL-to-IMU alignment and scale factor, and GNSS antenna lever arms at factory end-of-line, plus online calibration that keeps them accurate through shock, thermal cycling, and pressure.

• Onboard compute and real-time software: contribute to bringing up and running perception on embedded Linux (Yocto, GPU-accelerated inference with TensorRT or similar), processing multi-camera video, sonar, and hydrophone arrays in real time within tight power budgets.

• Core libraries: build the optimization, factor graph, and sensor-model libraries that other teams use for estimation, planning, and control.

• Logging, replay, and validation: build logging and deterministic reprocessing of mission data so every dive becomes a regression test, and drive reliability improvements from field results.

ABOUT YOU

REQUIREMENTS

• You've shipped perception or navigation software on autonomous vehicles, and you can show us what it does.

• Experience deploying state estimation on real vehicles in real environments — air, ground, marine, or space — outside of labs and simulation.

• Proven ability to take estimation systems from math to production code to field-validated performance.

• Strong foundations in state estimation: Kalman filtering, factor graphs and smoothing, nonlinear least-squares optimization, and observability analysis.

• Hands-on inertial navigation experience: IMU error modeling, preintegration, and tightly-coupled fusion with aiding sensors (GNSS, DVL, vision, or similar).

• Sensor calibration experience: camera intrinsics, extrinsics, and camera-IMU calibration, ideally both on the factory floor and online in the field.

• Proficiency in Rust or C++ for real-time, safety-critical software, and Python for prototyping and analysis.

• Experience with optimization and estimation frameworks (Ceres, GTSAM, SymForce, or similar) — or experience writing your own.

• Comfortable on embedded Linux compute: sensor drivers, time synchronization, and real-time performance on constrained hardware.

• Strong debugging instincts: you can trace a bad estimate back to timestamps, calibration, sensor faults, or math, using logs and replay.

• Experience building tools for logging, deterministic replay, and evaluation of autonomy software.

• Ability to collaborate across electrical and mechanical: sensor selection, mounting, time sync, and the hardware realities that shape perception.

• You have a track record of strong problem-solving skills and a demonstrated ability to work independently in dynamic environments.

NICE-TO-HAVES

• Marine navigation and sensing experience (DVL, USBL/SBL, INS, multibeam, sidescan, or forward-looking sonar)

• Acoustics and array processing: hydrophone or microphone arrays, beamforming, direction-of-arrival estimation

• Advanced GNSS: RTK, carrier-phase and time-differenced carrier-phase processing, correction services

• Deep learning perception deployed on edge hardware (TensorRT, ONNX), especially in degraded visibility

• Embedded firmware experience: microcontrollers, CAN, bootloaders, and low-level sensor interfaces

• Experience leading or mentoring a perception or navigation team

Think you're our next teammate? Apply and send examples of perception or navigation systems you've shipped or owned. Tell us what you built, how you knew it was working, what failed in the field, and how you made it reliable.