Здесь есть конкретная исследовательская задача на переднем крае и понятный путь от исследований к продукту, но зарплата заметно ниже ориентира для похожих ролей, что снижает привлекательность вакансии.
Ниже рынка
Сильная исследовательская задача
Современный стек
Перспективная сфера
?Много неопределенности
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Neural Objects Researcher (3D Reconstruction): Building neural systems that reconstruct manipulated objects from monocular egocentric video and turn them into simulatable 3D assets with an accent on occlusion, deformable objects, contact-aware reconstruction, and physical properties. Focus on designing open-ended research architectures, training deep-learning models on large-scale real data, and transferring prototypes into production through integrations.
Location: On-site in the Toronto GTA or New York
Compensation: CA$200K–CA$230K total wage range, plus potential equity, health and wellness benefits, and other company programs.
Company
hirify.global builds data infrastructure for robotics and embodied AI, including data capture, labeling, and hardware-enabled workflows for AI labs and robotics companies.
What you will do
Own the architecture for generating 3D objects from monocular and egocentric video of object manipulation.
Develop neural reconstruction methods for small, low-resolution, partially occluded, and deformable objects.
Use generative 3D methods to complete and refine partial reconstructions while rejecting hallucinated geometry.
Incorporate hand contact as a signal for constraining object shape and pose.
Extend reconstructed geometry toward simulatable assets with physical properties and rigging.
Build research prototypes, publish where appropriate, and hand results to the integrations researcher for productization.
Requirements
Strong research experience in neural implicit or explicit 3D, 4D reconstruction, or generative 3D.
Ability to design and train deep-learning models and read or reproduce frontier research papers.
Solid foundation in structure-from-motion, multi-view geometry, meshes, and differentiable rendering.
Research judgment for decomposing open-ended problems and making steady progress on real data.
Strong Python and PyTorch skills for rapid, clean research prototyping and model training.
Depth in 3D deep learning; the exact technology stack is less important than research capability.
Nice to have
Experience with diffusion or feed-forward generative 3D methods.
Experience with NeRF, Gaussian splatting, or differentiable rendering.
Experience with 4D or dynamic-scene reconstruction from video.
Experience with physics simulation, rigging, hand-object interaction, or mesh processing using Open3D or trimesh.
Publications at leading computer vision or graphics venues.
Culture & Benefits
Open-ended research role focused on an unsolved problem using real, large-scale egocentric data.
Clear path from research prototypes to production through a dedicated integrations role.
Direct impact on data used to train real robots and embodied AI systems.
Inclusive hiring practices and accommodation support throughout the interview process.
Potential equity, health and wellness benefits, and other company programs.