Research Intern – World-Action Model / VLA for Autonomous Driving

Bosch Group•Sunnyvale, CA

10d•$39 - $66

About The Position

The Bosch Research and Technology Center North America (RTC-NA) is part of the global Bosch Group and focuses on providing technologies and system solutions in areas like artificial intelligence, energy technologies, and advanced MEMS design. Our AI research in Silicon Valley concentrates on Foundation Models, Big Data Visual Analytics, Explainable AI (XAI), Natural Language Processing, Computer Vision & Mixed Reality, Cloud Robotics, Data Science, AI System Engineering, and Time-series Analysis. We develop scalable, intelligent, and trustworthy AIoT solutions for Bosch products and services in application areas such as automated driving, advanced driver assistance systems (ADAS), robotics, smart manufacturing, enterprise AI, health care, and smart home and building solutions. The Intelligent Autonomous Systems group, originating from this AI research, is responsible for enabling future autonomous Bosch products by advancing automated driving, ADAS, robotics, and automation through innovations in system architecture and AI components. This includes methods for motion planning, high-level task planning, decision making, and building frameworks for reliable distributed computing. We collaborate with internal Bosch business units and external academic and industry partners, aiming for top-venue publications and product impact.

Requirements

Ph.D. student in Computer Science, Robotics, Computer Engineering, or a related field (must be a currently enrolled student).
Minimum GPA of 3.0
Hands-on experience with at least one of: generative models (diffusion, autoregressive transformers, VAEs), world models / video prediction, or end-to-end autonomous driving.
Solid fine-tuning skills and proficiency with deep learning frameworks.
Experience with reinforcement learning and/or model-based planning.

Nice To Haves

Publication record in top venues including CVPR, ICCV, ECCV, ICLR, NeurIPS and ICML.
Familiarity with driving benchmarks such as NavSim and Bench2Drive.
Experience with large-scale distributed training and multi-modal / video data pipelines.
Able to work independently, with strong research and problem-solving skills.
Good communication and teamwork skills.

Responsibilities

Designing and training world-action models (autoregressive / diffusion-based) that predict future sensor observations or latent states conditioned on ego actions.
Integrating world models with planning for end-to-end or model-based driving policies.
Summarize research findings in high-quality paper and/or patent submissions.
Collaborate closely with other researchers, and strong work is expected to lead to top-venue publications and/or product impact.