Robotics Software Engineer - Learned Manipulation on Real Robots

Hivebotics; San Francisco (CA); Vor Ort; USD 150.000 - 230.000; Vollzeit

Zusammenfassung

Hivebotics in the United States builds autonomous robots for real commercial environments and leads the learning stack to enable new manipulation skills from real-world data.

You will own the learned manipulation capability end to end, focusing on grasping and contact-rich dexterous manipulation, taking tasks from concept to policies running on real hardware with rigorous experimentation.

Qualifikationen

• Degree in Robotics, CS, Software Engineering or equivalent practical experience.
• Hands-on experience with imitation learning from demonstrations.
• Working knowledge of open VLA model families (e.g. OpenVLA, Octo, π0, GR00T).
• Experience exporting and deploying models through ONNX or equivalent runtime.
• Experience with GPU-accelerated simulation (NVIDIA Isaac Sim/Lab).
• Strong Python skills and grounding in manipulation and control.

Aufgaben

• Build teleoperation and recording pipelines turning real deployments into training data.
• Fine-tune open-weight VLA base policies on our data using LoRA; build evaluation harnesses.
• Develop grasping and contact-rich manipulation policies and controllers on real hardware.
• Export policies to a portable runtime (ONNX) and deliver them to fleet deployment as versioned artifacts.
• Re-run base model selection on a regular schedule to adopt stronger foundation models.
• Document code, algorithms, experiments, and results clearly.

Kenntnisse

Robotics

Imitation learning

Python

Control theory

Experimentation

Ausbildung

Bachelor's degree in Robotics or related field

Tools

ONNX

NVIDIA Isaac Sim

Isaac Lab

LoRA

Jobbeschreibung

Hivebotics in the United States builds autonomous robots for real commercial environments and leads the learning stack to enable new manipulation skills from real-world data.

You will own the learned manipulation capability end to end, focusing on grasping and contact-rich dexterous manipulation, taking tasks from concept to policies running on real hardware with rigorous experimentation.