Job Description

Materials science and technology are our passion. With our cutting-edge research, Empa's around 1,100 employees make essential contributions to the well-being of society for a future worth living. Empa is a research institution of the ETH Domain.

Postdoctoral Researcher in Tabular Foundation Models for Building and District Energy Systems

Our passion lies in materials science and technology, and at the Urban Energy Systems Laboratory (UESL), we develop strategies and methods to support the creation of decarbonized, resilient, and equitable energy systems.

This PostDoc position is offered in collaboration with the Intelligent Maintenance and Operations Systems (IMOS) Laboratory at EPFL (Prof. Olga Fink). IMOS develops advanced machine-learning and AI methods for complex engineering and industrial systems, with a particular focus on improving their reliability, availability, and operational performance while enabling more efficient and cost-effective maintenance.

To advance the development of tabular foundation models for energy systems, we are seeking a highly motivated and skilled postdoctoral researcher. The project aims to develop foundation models that can learn from heterogeneous tabular data across buildings and district-scale energy systems and transfer across systems, operating conditions, and downstream tasks. The position combines Empa UESL’s expertise in developing and accessing energy system models with the methodological expertise of the IMOS Laboratory in machine learning and foundation models.

Your tasks

• Evaluate and benchmark existing pre-trained tabular foundation models for building- and district-scale energy applications, assessing their transferability and generalization across systems, operating conditions, and downstream tasks.
• Adapt and fine-tune existing foundation models for energy-system applications, investigating efficient adaptation strategies and the use of domain-specific data and knowledge.
• Develop new tabular foundation-model approaches where existing pre-trained models are insufficient, with a particular focus on transferability across heterogeneous energy systems and datasets.
• Validate and benchmark the developed models using building measurements, physics-based simulations and energy-system optimization models.
• Investigate how tabular foundation models can support energy-system modelling and optimization, including applications such as prediction, surrogate modelling, uncertainty quantification, and decision support.
• Coordinate the joint research activities between UESL and IMOS.
• Publish and present research perspectives and results.
• Contribute to research proposals and the acquisition of competitive funding.

Your profile

We seek a highly motivated and dedicated researcher with a PhD in mathematics, electrical or mechanical engineering, computer science or a related field, and a strong methodological background in machine learning. The ideal candidate has demonstrated research experience with foundation models, including the evaluation and adaptation of pre-trained models, fine-tuning strategies, and the development of new model architectures or learning approaches. Experience with tabular foundation models or foundation models for structured data is particularly relevant to this position.

Key qualifications include:

• Strong research experience in deep learning and foundation models, including experience with pre-trained models, fine-tuning, transfer learning, or self-supervised learning. Experience with tabular foundation models or foundation models for structured data is particularly relevant.
• A strong understanding of modern deep-learning architectures and training strategies, and experience designing and rigorously evaluating new machine-learning methods. • Excellent Python programming skills and strong hands-on experience implementing, training, and evaluating deep-learning models and research codebases.
• A strong track record in machine learning or closely related fields
• Excellent written and spoken English

Ideally, the candidate also:

• Has experience in energy system modeling and optimization
• Has experience with tabular or heterogeneous data, particularly across multiple da-tasets, domains, or tasks
• Is familiar with mathematical optimization methods, such as mixed-integer linear programming
• Has experience with uncertainty quantification, surrogate modelling or physics-informed machine learning
• Has an understanding of the technical challenges associated with the energy transition

Our offer

• A stimulating and interdisciplinary research environment and opportunities for personal and professional development
• Close collaboration between Empa and EPFL, bringing together complementary expertise in energy systems, machine learning, and foundation models
• The opportunity to shape an emerging research direction at the interface of artificial intelligence, energy-system modelling and optimization
• The possibility of starting immediately or by agreement

We foster a culture of inclusion and respect. We welcome all people who are interested in innovative, sustainable and meaningful activities - that's what counts.

We look forward to receiving your complete online application, including a letter of motivation, CV, a one-page research proposal on how you'd tackle your tasks, certificates, diplomas, and contact details for two referees. The research proposal should briefly outline the candidate’s research vision and proposed methodological directions for advancing tabular foundation models for energy systems, including potential research questions and approaches relevant to the position. Please submit your application exclusively via our job portal. Applications sent via email or post will not be considered.

Timeline

• Please submit your application by the 24th September 2026
• Interviews will take place throughout September until the 19th October 2026
• In case no suitable candidate can be found within the timeframe foreseen above, we will keep the position open. In that case, we are looking forward to your application after the above-mentioned deadline.