At Robot.com (formerly Kiwibot) we build AI-powered robots that help businesses automate processes, improve productivity, and free up human time for what truly matters. We believe the future will be powered by clean, efficient, and intelligent technological solutions.
Location : San Francisco, California (Hybrid)
About The Role
In this role, you'll push the frontier of visual‑language‑action models. You will play a vital role in our R&D cycle, working hands‑on with real‑world robotic systems. You will shape algorithms, data, training, and evaluation pipelines, turning novel ideas into models that power physical intelligence.
Responsibilities
- Designing and implementing experiments with robotic manipulation systems powered by vision‑language‑action models.
- Prototyping and evaluating robot foundation models for generalization across tasks and environments.
- Developing pipelines for data collection, training and fine‑tuning multimodal models using real‑world and simulated data.
- Collaborating with hardware, software, and ML teams to integrate research into deployable systems.
- Contributing to internal research publications, benchmarks, and open‑source tools.
- Supporting the deployment of intelligent behaviors in physical robots and simulation environments.
Requirements
Master’s, or PhD in Robotics, Computer Science, Electrical Engineering, or a related field. Strong programming skills in Python and C++, with experience in ML frameworks (e.g., PyTorch, JAX).Hands‑on experience with robotic systems, especially in manipulation or mobile platforms.Familiarity with vision‑language models (e.g., CLIP, Gemma, VIMA, RT‑2) and their application to robotics.A research‑oriented mindset with the ability to design, run, and analyze experiments.Bonus Points For
Experience training or fine‑tuning large‑scale foundation models.Publications in top‑tier robotics or ML conferences (e.g., ICRA, CoRL, NeurIPS, CVPR).Experience with ROS / ROS2 and simulation tools (e.g., Gazebo, Isaac Sim, MuJoCo).Familiarity with reinforcement learning, imitation learning, or self‑supervised learning in robotics.Contributions to open‑source robotics or ML libraries.#J-18808-Ljbffr