About this role
We're building the next generation of computer vision technology for wearables. You'll own a core technical area spanning reconstruction and generative enhancement, and drive it from research through production on wearables.
Responsibilities
Own a core area of our reconstruction stack and set its technical direction
Advance 3D visual representations, optimizing them for on-device delivery and representation
Develop image composition and correction techniques
Lead generative quality improvement for real-world capture
Optimize models for on-device inference under hard compute, memory, latency, thermal, and power constraints, and partition work between device and cloud
Define image quality gates
Own training data strategy,
Partner with product and platform teams to take prototypes through dogfooding and experimentation to shipped features
Shape the technical roadmap for your area and influence it across organizational boundaries
Deliver production-quality code, review others' work, and raise the engineering bar
Qualifications
Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience
PhD in Computer Science, Electrical Engineering, Robotics, Applied Math, or a related field with a computer vision, graphics, or ML focus — or equivalent practical experience
6+ years building production computer vision or graphics systems
Deep expertise in geometric vision: SfM, SLAM/VIO, multi-view stereo, calibration, bundle adjustment, or learned depth
Proven experience training and deploying deep learning models (PyTorch or equivalent)
Expert-level C++ and strong Python
Demonstrated record of taking research from prototype to shipped product Hands-on experience with 3D scene representations
Track record of technical leadership setting direction, mentoring, and driving multi-team efforts
Experience with generative image and video models
Strong publication record at top-tier venues (CVPR, ICCV, ECCV, NeurIPS, SIGGRAPH, 3DV)
Experience with camera ISP pipelines, computational photography, or multi-sensor fusion