Job Description
About the Role
We are seeking a versatile, hands-on Computer Vision Engineer to lead the development of both real-time spatial perception and post-processing 3D reconstruction systems for MOBY's subsea vehicles.
You will combine classical computer vision, modern vision models, and embedded deployment to build algorithms that run in real time on constrained hardware, working through non-uniform lighting, turbidity, marine life, and low-contrast seafloor.
Key Responsibilities
Real-Time Perception & Edge Navigation
• Build, optimize, and deploy real-time perception pipelines for obstacle detection, tracking, and hazard avoidance on embedded GPU hardware
• Develop robust Visual SLAM and state-estimation pipelines tuned for feature-sparse, visually degrading underwater environments
• Write highly efficient runtime code (C++/CUDA/TensorRT) maximizing frame rates while remaining compute- and memory-bounded on edge hardware
3D Spatial Reconstruction & Post-Processing
• Design and implement high-fidelity offline 3D reconstruction pipelines using multi-view stereo, 3D Gaussian Splatting, NeRFs, or structure-from-motion (SfM)
• Develop semantic tagging, object classification, and spatial map segmentation algorithms using state-of-the-art vision and vision-language models (VLMs/VLAs)
• Design camera calibration workflows (intrinsic, extrinsic, and multi-camera stereo) robust to refractive underwater optics and lens distortion
Hardware Prototyping & Field Integration
• Collaborate directly with hardware and embedded systems engineers to integrate cameras, lighting synchronization (PWM/hardware triggering), and IMU sensor fusion
• Rapidly prototype, benchmark, and field-test vision algorithms on real subsea crawler/ROV/AUV video feeds and datasets
Requirements
Required Qualifications
• Experience & Education: Master's degree with 2+ years of relevant industry experience, or a Bachelor's degree with 4+ years of relevant experience
• Core CV & Math: Deep understanding of classical computer vision (epipolar geometry, camera models, image restoration, feature extraction) and modern deep learning
• Underwater Visual Domain: Demonstrated experience handling subsea vision challenges (refractive optics, light attenuation, backscatter, non-uniform illumination, and low-contrast marine terrain)
• SLAM & State Estimation: Hands-on experience with Visual-Inertial SLAM (e.g., ORB-SLAM3, OKVIS, VINS-Mono) or sensor fusion algorithms
• Embedded Deployment & Compute Optimization: Proven capability deploying models to edge platforms (e.g., NVIDIA Jetson / CUDA / TensorRT / ONNX Runtime) with strict latency and memory budgets
• Practical Prototyping Skills: Strong proficiency in Python and C++, with hands-on experience using OpenCV, PyTorch/LibTorch, ROS 2, and point cloud libraries (PCL/Open3D)
Preferred Qualifications
• Experience with 3D Gaussian Splatting, Neural Radiance Fields (NeRFs), or dense photogrammetry pipelines
• Familiarity with hardware-triggered multi-camera synchronization and subsea lighting controls
• Exposure to Vision-Language Models (VLMs) or vision-guided manipulation and grasping
• Experience processing multi-modal sensor streams (DVL, IMU, sonar, monocular/stereo RGB)
Benefits
What We Offer
• Competitive pay
• Direct ownership of a critical, foundational platform in deep-sea robotics
• Small, highly technical team with broad systems exposure
• Rapid iteration between algorithm design, field testing, and real subsea deployment
• Opportunity to shape next-generation ocean intelligence infrastructure
