SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mecka AI is building the data infrastructure layer for robotics and embodied AI. The company designs and operates global systems for data capture, labeling, and hardware-enabled workflows used by leading AI labs and robotics companies to train and validate humanoid and embodied AI systems.
You will lead the development of a full-body motion estimation system that bridges egocentric camera observations with complete, physically plausible human pose. Currently, the pipeline estimates hand and upper-body keypoints from head-mounted stereo/mono cameras and ego-exo multi-camera volumes, but infers the lower body and feet using simplistic assumptions (ankle-plus-offset on flat floor, contact via frame removal). Your work will replace this with a learned full-body motion prior that conditions on partial evidence—metric camera trajectory, hand/wrist poses, upper-body keypoints, 2D leg detections, and full-body ground truth from the ego-exo volume—to produce metric world-frame poses that respect physics: no foot skate, no penetration, proper ground contact, and plausible balance including floor work (kneeling, sitting, lying, yoga).
Key responsibilities:
- Train generative motion models (diffusion, autoregressive, or masked) on mocap-scale and proprietary rig data, conditioned on partial egocentric evidence, to synthesize missing lower-body and feet.
- Implement physics-based tracking or contact-aware optimization to ensure kinematic estimates are physically consistent and grounded.
- Develop per-frame foot and body contact estimation with calibrated confidence alongside pose output.
- Work with body models (MHR, SOMA-X) to ensure production-ready output and pose priors that downstream solvers can consume.
- Collaborate with the audit team to define lower-body, feet, and contact metrics; commission motion captures (ego-exo, mocap, IMU suit) to close named gaps.
- Prototype → production: move from research to a runnable model in a shared repository that production owners integrate; publish findings.
You will have access to hundreds of hours of ego-exo multi-view ground truth, thousands of hours of egocentric recordings, and on-demand mocap and IMU-suit sessions. Success is measured by lower-body and feet pass rates, physical-violation metrics (penetration, foot skate, ground-plane errors) that improve each quarter, robust floor-work handling, and a public benchmark result and publication within the first year.
This is a named seat that sets the body pod's direction, with direct impact on the data that trains real robots and publication rights.
REQUIREMENTS:
- Research record in physics-based character animation, human motion generation, or 3D human pose and shape: PhD or equivalent body of work with publications at SIGGRAPH/SIGGRAPH Asia, CVPR/ICCV/ECCV, NeurIPS/ICLR, or CoRL.
- Hands-on experience training diffusion, autoregressive, or masked motion models on mocap-scale data with conditioning (keyframes, end effectors, paths, text).
- Proficiency in physics-based control: motion imitation and tracking for simulated humanoids (DeepMimic, AMP, MaskedMimic lineage); comfort with Isaac Gym/Isaac Lab or MuJoCo and reinforcement learning at scale.
- Knowledge of body models and kinematics: SMPL-X, MHR, or SOMA-X; retargeting; contact and collision modeling.
- Clean PyTorch research code and ability to work with noisy, real, large-scale data rather than clean benchmarks.
- Research taste: ability to decompose "full body from a head camera" into measurable steps and make steady progress.
STRONG PLUS:
- Egocentric or partial-observation full-body estimation (head- and hand-tracking to body).
- Human–scene interaction and contact; scene-aware motion synthesis; volumetric body models.
- Motion cleanup or denoising of estimated or corrupted motion.
- Reusable motion priors for control, generative controllers, or large-scale motion tokenization.
- Shipped research models into products or game engines; familiarity with Momentum/MHR tooling.
