Sieve
Member of Technical Staff, Robotics Research
San Francisco · On-site · Staff · Full-time · Posted Oct 23, 2026
- Experience
- 4+ years of research or engineering experience in robot learning, embodied AI, or 3D computer vision.
- Education
- A PhD is welcome but not required; strong publications, open-source projects, or systems the candidate can explain are also valued.
About the job
Lead research on how multimodal and 3D data affects robot learning by training and evaluating world-action models, building 3D labeling systems, and turning results into datasets and capture protocols.
What you'll do
- Run data ablations and train or post-train robot policies and action-conditioned world models; evaluate them on unseen tasks and data.
- Build 3D-from-video pipelines for geometry, pose, tracking, and scene reconstruction, and use simulation where useful for evaluation or augmentation.
- Convert findings into dataset curation and capture practices, collaborate with partner labs, and share results through publications and other technical outputs.
What they ask for
- 4+ years of research or engineering experience in robot learning, embodied AI, or 3D computer vision.
- Fluent Python and PyTorch or JAX; able to write, debug, and modify training code.
- Hands-on 3D vision experience, such as multi-view geometry, depth, pose, tracking, or reconstruction.
- Have trained a policy or world model and tested it on data not seen during training.
Nice to have
- Robot learning from egocentric or human video, real-robot closed-loop evaluation, or experience with MuJoCo or Isaac Sim.
- Large-scale distributed training or experience as an early startup hire.
Tools
Python, PyTorch, JAX, MuJoCo, Isaac Sim
AI focus
Train and evaluate world-action models, robot policies, and 3D perception systems; study how data choices affect robot capabilities.
Model-written summary · the original posting is authoritative