Build intelligence for the physical world.

Why Origami

Join a small, hands-on team building general-purpose dexterous manipulation. You’ll work across robot learning, controls, software, and hardware, test ideas quickly on physical robots, and turn failures into better data, models, and systems. Your work will directly shape what our robots can do.

In This Role You Will

• Train, deploy, and improve dexterous manipulation policies on physical robots.
• Build teleoperation and data-collection workflows for training and evaluation.
• Apply imitation learning and RL fine-tuning where appropriate.
• Build custom real-time control and state estimation algorithms, integrate learned policies with perception and planning, and debug failures across the full robotic system.

What We Hope You’ll Bring

• You have trained and deployed a learned policy on a physical robot.
• Strong Python and hands-on experience building and debugging robotic systems.
• Experience with PyTorch or JAX and methods such as imitation learning or RL.
• Working knowledge of robot kinematics, controls, and motion planning.
• Strong experimental judgment and ownership from idea to real-world demonstration.

Bonus Points If You Have

• Experience with dexterous or bimanual robots, including tactile or force sensing.
• Experience shipping real robotic systems using ROS2, C++, sim-to-real, or edge inference.
• Designed and built a robot or major subsystem through FTC, FRC, Formula SAE, Formula Hybrid, or a comparable project.

Why Origami

Join a small team building the software foundation for general-purpose dexterous manipulation. Your systems will power every robot rollout, training run, and physical evaluation, giving researchers the tools to move faster and making each experiment compound into better robot behavior.

In This Role You Will

You will own critical parts of the learning pipeline and make every experiment reproducible, every model traceable, and every real-world rollout useful.

• Own the pipeline from teleoperated robot rollouts to calibrated, synchronized, and versioned multimodal datasets, with tools for replay, visualization, curation, and failure mining.
• Build reproducible training infrastructure for launching jobs, tracking experiments, managing checkpoints and model lineage, and connecting results back to source data.
• Build physical evaluation systems with repeatable task runners, held-out tests, result logging, and dashboards.

What We Hope You’ll Bring

• Strong Python and comfort working in C++.
• Experience building reliable data, ML, or distributed systems used by other engineers.
• Experience designing observable, reproducible workflows for multimodal or time-series data.
• Ability to own and debug systems spanning robot software, sensors, networking, storage, and compute.

Bonus Points If You Have

• Experience building robot-learning or dexterous-manipulation infrastructure with tools such as ROS2, MCAP, Protobuf, Foxglove, Ray, Slurm, Parquet, or Arrow.
• Expertise in controls, low-latency software, sensor synchronization, or real-time systems.
• Built and operated robots through FTC, FRC, Formula SAE, Formula Hybrid, an autonomous vehicle team, or a substantial personal robotics project.

Why Origami

Join a small team teaching robots to perform useful dexterous tasks in the physical world. You’ll own a focused project, work directly with engineers across learning, controls, and hardware, and test your ideas on the same robots used by the full-time team.

In This Role You Will

• Collect and curate manipulation demonstrations.
• Train imitation-learning, diffusion-policy, or RL models.
• Run physical evaluations and analyze model and hardware failures.
• Build visualization, debugging, and policy-evaluation tools.
• Integrate learned policies with perception and control.

What We Hope You’ll Bring

• Current or recent BS, MS, or PhD student in a relevant technical field.
• Strong Python and experience with a deep-learning framework.
• You have trained a meaningful model and worked with a physical robot.
• Evidence of curiosity and independent building through a lab, team, competition, or personal project.

Bonus Points If You Have

• Experience with imitation learning, reinforcement learning, diffusion policies, or VLA models.
• Experience with manipulation, teleoperation, ROS2, or tactile and force sensing.
• A project you can demonstrate through code, video, or a clear technical write-up.

Why Origami

Join a small team building the platform that turns physical robot experience into better behavior. You’ll own a focused infrastructure project used by robotics and ML engineers, work on real systems, and ship something that lasts beyond the internship.

In This Role You Will

• Build a robot episode viewer and failure-annotation interface.
• Automate dataset validation and quality reporting.
• Create a reproducible training launcher with experiment tracking.
• Implement an on-robot task-evaluation harness.
• Trace dataset lineage from physical rollout to trained checkpoint.
• Develop failure-mining and targeted data-generation tools.

What We Hope You’ll Bring

• Current or recent degree candidate in robotics, computer science, or a related field.
• Strong Python and experience with C++ or robotics systems.
• You have trained or deployed at least one ML model.
• Experience with a physical robot and practical debugging.
• Evidence of independently building difficult technical projects.

Bonus Points If You Have

• Experience building data pipelines, training tools, experiment tracking, or evaluation harnesses.
• Familiarity with ROS2, multimodal robot data, cloud compute, or containerized workloads.
• A tool or system that other researchers or engineers have used.