Master's thesis: on Distributed and Federated AI for Autonomous Vehicles

The volume and diversity of sensor data produced by autonomous vehicles create communication, storage and privacy challenges for centralized processing. Within the DREAM project, these two Master's thesis topics explore federated learning for real-time vehicle applications. They focus on learning from unlabeled driving data and transferring knowledge between heterogeneous AI models as vehicle platforms change.

The assignment

• Topic 1: Design and test self-supervised learning methods for driving data, explore combinations of supervised and self-supervised federated learning, and benchmark against relevant supervised and federated approaches.
• Topic 2: Explore knowledge distillation and transfer between heterogeneous models, including adaptation to changes in architectures, sensors or hardware, and assess these methods in a federated learning setting.
• Implement and evaluate methods experimentally with autonomous-driving datasets, including the Zenseact Open Dataset (ZOD).
• Agree on the final research scope with the supervisors in line with DREAM's objectives and present findings to the project partners.

Who we are looking for

• Master's students with strong knowledge of machine learning and computer vision.
• Skills in deep learning and Python, experience with complex systems, and the ability to understand scientific publications.
• Experience in federated learning, self-supervised learning, knowledge distillation or autonomous-driving data is an advantage.

What we offer

• Close research collaboration with RISE and Zenseact.
• Supervision by Sima Sinaei, Henrik Abrahamsson and Mina Alibeigi.
• Compensation of SEK 39,000 upon successful completion of a high-quality thesis.

Practical details

Two Master's students are sought for two thesis topics, each worth 30 credits. Start date: 11 January 2027. Location: RISE's Kista office in Stockholm, with on-site attendance required at least three days per week. Apply with a CV, motivation letter and transcript of records by 15 December 2026. Applications are assessed continuously, so early submission is encouraged. Questions: Henrik Abrahamsson, henrik.abrahamsson@ri.se, or Sima Sinaei, sima.sinaei@ri.se.