About
"Any sufficiently advanced technology is indistinguishable from magic" --Arthur C Clark
Staff Machine Learning Research Scientist
Hi, I am Jonathan. I currently work as a Staff Machine Learning Research Scientist at Nuro where I focus on solving complex problems at the intersection of Prediction, Planning, Decision Making, and Simulation. My interests lie in the research, development, and application of techniques from machine learning and reinforcement learning to solve some of the most challenging problems.
In my current role, I am developing reinforcement learning algorithms for solving planning in the autonomous driving space. Previously, I have tackled the problems of learning from large-scale and diverse-quality demonstrations from crowd sourcing as well as reinforcement learning in high-dimensional action spaces for humanoid robotics.
If any of these topics are interesting to you, feel free to reach out! I am always happy to talk.
- City: Mountain View, CA
- Email: jaustinb1@gmail.com
Resume
Professional Experience
Staff Machine Learning Research Scientist
2024 - Present
Nuro, Mountain View, CA
- Leading development of reinforcement learning and related methods for various different applications such as ego planning and agent behavior modeling. Techniques include world modeling, RLHF, reward learning, model based and model free approaches, safe RL, etc.
Senior Machine Learning Research Scientist
2022 - 2024
Nuro, Mountain View, CA
- Leading the design, development, and integration of sophisticated machine learning models (generative modeling, reinforcement learning, etc) to generate and select ego motion plans to improve decision making on the road.
- Developing technical roadmaps and driving execution of large cross functional projects.
Machine Learning Engineer
2021 - 2022
Nuro, Mountain View, CA
- Developed novel machine learning solutions to complex sequential decision-making problems in autonomous driving applications.
- Created and optimized large scale distributed training systems and efficient data pipelines.
- Drove efforts to integrate new ML models into an existing software stack.
Member of the Technical Staff
2018 - 2020
Giant.ai, Campbell, CA
- Worked directly with the founders to develop reinforcement learning models and infrastructure for humanoid robotics.
- Developed imitation and curriculum learning methods for learning high dimensional (60-dim) continuous control policies.
- Created robust robotics controllers to simplify the control scheme of the robot.
Software Engineering Intern
2019
Google, Mountain View, CA
- Designed and developed memory and speed optimized neural networks for integration into the AV1 video compression codec.
- Developed methods to efficiently utilize neural networks for complex rate-control optimization problems.
- Integrated and tested my framework within the broader system being developed.
Education
Master of science: Computer Science
2020 - 2021
Stanford University
Graduate degree in computer science with a focus on Artificial Intelligence. Coursework including: reinforcement learning, robotics, and NLP.
Bachelor of Science: Computer Science
2016 - 2020
Stanford University
Undergraduate degree in computer science with a focus on Artificial Intelligence. Coursework including: reinforcement learning, computer vision, robotics, optimization, and NLP.
Research Experience
Stanford Vision And Learning Lab
2018 - 2021
Stanford University
Developed RoboTurk, the first system to enable crowd sourcing of large-scale robotics datasets using teleoperation in simulation and in the real world. Helped design novel methods for learning from diverse and multi-modal human demonstrations including methods in goal directed reinforcement learning, behavior cloning, reward learning, and curriculum learning. Work published in CoRL 2018 and IROS 2019.
Publications
CIMRL: Combining IMitation and Reinforcement Learning for Safe Autonomous Driving
2024
Arxiv
Booher, Jonathan, Khashayar Rohanimanesh, Junhong Xu and Aleksandr Petiushko. “CIMRL: Combining IMitation and Reinforcement Learning for Safe Autonomous Driving.” (2024).
ArxivMulti-Constraint Safe RL with Objective Suppression for Safety-Critical Applications
2024
Arxiv
Zhou, Zihan, Jonathan Booher, Wei Liu, Aleksandr Petiushko, and Animesh Garg. "Multi-Constraint Safe RL with Objective Suppression for Safety-Critical Applications." arXiv preprint arXiv:2402.15650 (2024).
ArxivRoboTurk: A Crowdsourcing Platform for Robotic Skill Learning through Imitation
2018
Presented at CoRL
Ajay Mandlekar, Yuke Zhu, Animesh Garg, Jonathan Booher, Max Spero, Albert Tung, Julian Gao, John Emmons, Anchit Gupta, Emre Orbay, Silvio Savarese, & Li Fei-Fei (2018). RoboTurk: A Crowdsourcing Platform for Robotic Skill Learning through Imitation. CoRR, abs/1811.02790.
ArxivScaling Robot Supervision to Hundreds of Hours with RoboTurk: Robotic Manipulation Dataset through Human Reasoning and Dexterity
2019
Presented at IROS
Ajay Mandlekar, Jonathan Booher, Max Spero, Albert Tung, Anchit Gupta, Yuke Zhu, Animesh Garg, Silvio Savarese, & Li Fei-Fei (2019). Scaling Robot Supervision to Hundreds of Hours with RoboTurk: Robotic Manipulation Dataset through Human Reasoning and Dexterity. CoRR, abs/1911.04052.
Arxiv