Research Scientist (Human Motion Generation)

Location: On-site in Toronto GTA or New York

Salary: CA$200K–CA$230K per year, 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 training and validating humanoid and embodied AI systems.

What you will do

• Build a full-body motion model that reconstructs pose in a metric world frame from egocentric camera trajectories, hand and wrist poses, upper-body keypoints, partial leg detections, and multi-view ground truth.
• Train generative motion models, including diffusion, autoregressive, or masked models, to infer unseen lower-body and foot motion.
• Develop physics-based tracking, imitation, or contact-aware optimization to prevent penetration, foot skating, sinking, and implausible balance.
• Produce per-frame foot and body contacts, ground-plane estimates, and calibrated confidence alongside the pose.
• Define lower-body, foot, and contact metrics with the audit team and commission mocap, IMU-suit, and ego-exo captures for benchmark gaps.
• Prototype and deliver a runnable model for production integration, with opportunities to publish research results.

Requirements

• Research record in physics-based character animation, human motion generation, or 3D human pose and shape, supported by a PhD or equivalent body of work.
• Publications or equivalent research contributions in venues such as SIGGRAPH, CVPR, ICCV, ECCV, NeurIPS, ICLR, or CoRL.
• Experience training generative motion models on mocap-scale data and conditioning them on constraints such as keyframes, end effectors, paths, or text.
• Experience with physics-based control, motion imitation, simulated humanoids, reinforcement learning at scale, and frameworks such as Isaac Gym, Isaac Lab, or MuJoCo.
• Experience with SMPL-X, MHR, or SOMA-X, including retargeting, contact modeling, and collision modeling.
• Strong Python and PyTorch engineering skills, with experience working on noisy, real-world, large-scale data.

Nice to have

• Experience with egocentric or partial-observation full-body estimation.
• Experience with human–scene interaction, contact-aware motion synthesis, volumetric body models, motion cleanup, or denoising.
• Experience shipping a research model into a product or game engine, or familiarity with Momentum and MHR tooling.
• Experience with ProtoMotions- or MimicKit-class physics-based character-control frameworks and AMASS-class motion capture.

Culture & Benefits

• Access to hundreds of hours of ego-exo multi-view ground truth, thousands of hours of egocentric recordings, mocap, and IMU-suit sessions.
• Ownership of an open research problem with a defined research seat and production owners for integration.
• Publication rights and compute resources are provided.
• Work directly on data infrastructure that trains real robots.
• Inclusive hiring practices and interview accommodations are available on request.