Doctoral student in physics-guided foundation model for time-series data

Join Chalmers to develop physics-guided, data-driven foundation models for multivariate time-series in safety-critical systems, with a primary focus on automotive applications (REF 2026-0434). Design predictive and generative models, scale training on real data, and validate in simulation and with industry partners.

About us

The Department of Computer Science and Engineering is a joint department of Chalmers and the University of Gothenburg. At the Division of Computing Science, research spans secure and trustworthy software and systems.

This project is a collaboration between AIXLab@Chalmers and Volvo Group. You join AIXLab (usable AI for real-world settings) and work closely with Volvo Group engineers and researchers, with access to industrial datasets, simulation environments, and validation workflows.

About the research project

The project advances physics-aware foundation models—reusable, pretrained time-series models adaptable across vehicles, conditions, and tasks. Primary use case: automotive prediction of vehicle behavior, simulation of rare safety-critical scenarios, and test-case generation. Techniques are designed to transfer to other safety-critical domains such as healthcare.

Work focuses on multivariate vehicle time series (CAN signals, sensor streams, simulated trajectories) with models that integrate physical structure (dynamics, constraints, conservation laws) into large neural architectures via constraints, inductive biases, hybrid simulation–learning loops, or physics-consistent losses. Combines forecasting, representation learning, and scenario generation under safety and reliability constraints.

Who we are looking for

Mandatory: Master's (masterexamen 120 / magisterexamen 60; 4-year Bachelor's outside Sweden accepted) in Computer Science, Electrical engineering, or equivalent; strong English; strong ML fundamentals and interest in time-series and physics-guided ML; proficiency in Python and modern DL frameworks (e.g. PyTorch); engineering maturity for large-scale GPU/cluster training, reproducible pipelines, versioned datasets, and systematic evaluation; ability to formulate research questions and run empirical studies at scale.

Strengthening: physics-informed ML; foundation models for time-series; safety-critical systems/scenario generation; academic research and publications.

What you will do

• Advanced courses in the Graduate school of Computer Science and Engineering
• Develop scientific concepts and communicate results
• By the end of the PhD, deliver reusable modeling frameworks for safety-critical time-series data, publications in top-tier venues, and contributions influencing industrial validation pipelines

Contract terms

• Fully funded from start
• Limited to four years; teaching up to 20% may extend to five years
• Starting salary 34,550 SEK per month (valid from 25 May 2025)
• Physical presence required; valid residence permit by study start

Application procedure

English PDF application via the official Chalmers vacancy page: CV, personal letter, Bachelor's and if available master's thesis with transcripts. Deadline: 1 October 2026. Research questions: Yinan Yu. Application process: Carl-Johan Seger.

Mandatory

• Master's degree (masterexamen 120 credits or magisterexamen 60 credits) in Computer Science, Electrical engineering, or equivalent (4-year Bachelor's accepted for education earned outside Sweden)
• Strong written and verbal English
• Strong machine learning fundamentals (probability, statistics, optimization) and strong interest in time-series modeling and physics-guided machine learning
• Proficiency in Python and modern deep learning frameworks (e.g. PyTorch)
• Engineering maturity: large-scale GPU/cluster training, reproducible experiment pipelines, versioned datasets, systematic evaluation
• Ability to formulate research questions and run empirical studies at scale

Strengthening

• Experience with physics-informed machine learning
• Background in foundation models for time-series (forecasting, representation learning)
• Exposure to safety-critical systems, scenario generation, or test coverage for edge cases
• Experience with academic research and publications