Internship / Masters Thesis - Sim‑to‑Real Transfer in Reinforcement Learning (f/m/d)
Internship / Masters Thesis - Sim‑to‑Real Transfer in Reinforcement Learning (f/m/d)
Mönsheim, DE, 71297 München, DE, 80807 Berlin, DE, 10587
Internship / Masters Thesis - Sim‑to‑Real Transfer in Reinforcement Learning (f/m/d)
We are CARIAD, the automotive software company of the Volkswagen Group. Our teams build automotive software platforms and digital customer functions for iconic brands like Audi, Volkswagen, and Porsche – supporting the Volkswagen Group in becoming the leading automotive technology company. With CARIDIANS in Germany, the USA, China, Estonia, and India, we are transforming automotive mobility for everyone.
Join us and be part of this exciting journey!
YOUR TEAM
For the department Vehicle, Energy, Motion & Body (VEMB) we are looking for a student (intern or master thesis) to work on a research topic in the area of reinforcement learning and simulation‑to‑real transfer. Our department develops advanced software solutions for vehicle energy, motion, and body systems. Within VEMB, our pre‑development team focuses on learning‑based methods for control and decision‑making, aiming to enable faster, scalable, and more cost‑effective development of onboard functions. A central challenge in this context is the reliable transfer of reinforcement learning policies trained in simulation to real systems, which requires systematic approaches to handle model uncertainties and real‑world variability.
WHAT YOU WILL DO
- Work together with a PhD student in the field of reinforcement learning and sim‑to‑real transfer
- Review the state of the art in domain randomization, adaptive reinforcement learning, and policy transfer
- Investigate advanced domain randomization techniques to improve robustness and real‑world performance of simulation‑trained reinforcement learning policies
- Use real‑world measurement data to reduce the simulation‑to‑reality gap by tuning, adapting, or constraining simulation models
- Design and conduct experiments to systematically evaluate the impact of different randomization and adaptation strategies
- Assist in implementing prototype learning pipelines and validate developed methods in simulation and selected real‑world experiments
- Collaborate with teams in pre‑development and series development environments
WHO YOU ARE
- Enrolled student in a relevant field such as Computer Science, Robotics, Electrical Engineering, or Mechatronics, with a strong focus on machine learning
- Strong foundation in machine learning and reinforcement learning, including a solid understanding of modern learning algorithms and training paradigms
- Solid programming skills in Python and hands‑on experience with modern ML frameworks (preferably JAX)
- Experience with designing, training, and evaluating learning‑based models in simulation environments
- Basic understanding of control systems, simulation, or physical modeling is a plus
- Structured and independent working style with strong analytical and problem‑solving skills
- Fluency in English and German and good communication skills
NICE TO KNOW
- Remote work options within Germany
- Duration: 6 months
- 35-hour week
- Salary: 13,90 €/hour
At CARIAD, we embrace individuality and diversity because we believe our differences make us stronger. We actively seek to build teams with a variety of backgrounds, perspectives, and experiences. Our goal is to create an environment where everyone feels valued and empowered to contribute. If you need assistance with your application due to a disability, please reach out to us at careers@cariad.technology - we are happy to support you.
We are CARIAD, the automotive software company of the Volkswagen Group. Our teams build automotive software platforms and digital customer functions for iconic brands like Audi, Volkswagen, and Porsche – supporting the Volkswagen Group in becoming the leading automotive technology company. With CARIDIANS in Germany, the USA, China, Estonia, and India, we are transforming automotive mobility for everyone.
Join us and be part of this exciting journey!
YOUR TEAM
For the department Vehicle, Energy, Motion & Body (VEMB) we are looking for a student (intern or master thesis) to work on a research topic in the area of reinforcement learning and simulation‑to‑real transfer. Our department develops advanced software solutions for vehicle energy, motion, and body systems. Within VEMB, our pre‑development team focuses on learning‑based methods for control and decision‑making, aiming to enable faster, scalable, and more cost‑effective development of onboard functions. A central challenge in this context is the reliable transfer of reinforcement learning policies trained in simulation to real systems, which requires systematic approaches to handle model uncertainties and real‑world variability.
WHAT YOU WILL DO
- Work together with a PhD student in the field of reinforcement learning and sim‑to‑real transfer
- Review the state of the art in domain randomization, adaptive reinforcement learning, and policy transfer
- Investigate advanced domain randomization techniques to improve robustness and real‑world performance of simulation‑trained reinforcement learning policies
- Use real‑world measurement data to reduce the simulation‑to‑reality gap by tuning, adapting, or constraining simulation models
- Design and conduct experiments to systematically evaluate the impact of different randomization and adaptation strategies
- Assist in implementing prototype learning pipelines and validate developed methods in simulation and selected real‑world experiments
- Collaborate with teams in pre‑development and series development environments
WHO YOU ARE
- Enrolled student in a relevant field such as Computer Science, Robotics, Electrical Engineering, or Mechatronics, with a strong focus on machine learning
- Strong foundation in machine learning and reinforcement learning, including a solid understanding of modern learning algorithms and training paradigms
- Solid programming skills in Python and hands‑on experience with modern ML frameworks (preferably JAX)
- Experience with designing, training, and evaluating learning‑based models in simulation environments
- Basic understanding of control systems, simulation, or physical modeling is a plus
- Structured and independent working style with strong analytical and problem‑solving skills
- Fluency in English and German and good communication skills
NICE TO KNOW
- Remote work options within Germany
- Duration: 6 months
- 35-hour week
- Salary: 13,90 €/hour
At CARIAD, we embrace individuality and diversity because we believe our differences make us stronger. We actively seek to build teams with a variety of backgrounds, perspectives, and experiences. Our goal is to create an environment where everyone feels valued and empowered to contribute. If you need assistance with your application due to a disability, please reach out to us at careers@cariad.technology - we are happy to support you.
Mönsheim, DE, 71297 München, DE, 80807 Berlin, DE, 10587