Doctoral researcher - Uncertainty-Aware ML Force Fields

The University of Luxembourg invites applications for a fully funded Ph.D. position in machine-learning force fields (MLFFs), uncertainty quantification, and atomistic simulations within the FNR-funded UMLFF project.

MLFFs have transformed atomistic simulations, offering quantum-chemical accuracy for large systems. However, they currently lack reliable uncertainty estimates, limiting error detection and automation. The UMLFF project aims to develop next-generation MLFFs with built-in uncertainty predictions to enable safe, automated active learning and create broad, reliable MLFFs.

You will join the Theoretical Chemical Physics group led by Prof. Alexandre Tkatchenko, supervised by Dr. Igor Poltavskyi.

Your research tasks will include:

• Uncertainty Estimation in Deep Neural Networks for MLFFs
  • Implement and test uncertainty-aware loss functions
  • Study calibration and post-calibration for predictive uncertainty
  • Integrate uncertainty modules into MLFF architectures
• Detecting Extrapolation and Low-Reference Regimes
  • Analyze MLFF feature spaces to find out-of-distribution atomic environments
  • Use chemical neighborhood representations to detect sparse or unseen cases
  • Combine ML features with chemical descriptors to enhance uncertainty robustness
• Building Uncertainty-Aware General Purpose Models
  • Merge developments into a unified, efficient MLFF architecture
  • Train MLFFs on large, diverse datasets across broad chemical space
  • Evaluate models through molecular dynamics, simulations, and benchmarks
• Active Learning in Configurational and Chemical Spaces
  • Integrate uncertainty-aware MLFFs into active learning frameworks
  • Explore automated dataset generation for molecules and materials
  • Contribute to open-source tools for automated MLFF training

Your profile

Essential qualifications:

• Master's degree in Physics, Chemistry, Materials Science, Computer Science, or related fields
• Strong background in theoretical/computational physics or chemistry
• Experience with machine learning, Python, and modern ML frameworks like PyTorch/JAX
• Genuine interest in MLFFs, simulation methods, and foundational ML research

Desired skills:

• Experience with atomistic simulation codes: ASE, FHI-aims, VASP, CP2K
• Knowledge of deep learning architectures, graph neural networks, or uncertainty quantification
• Familiarity with HPC environments

Language Requirements: Applicants must demonstrate at least B2-level proficiency in the language of their thesis. For details and accepted certificates, please visit the Application for admission - Doctoral Candidates.

How to apply

Please prepare a single PDF file including:

• Curriculum Vitae, including relevant coursework, ML experience, and publications, if any
• Cover letter presenting your motivation for this doctoral thesis topic, and explaining how your qualifications and aspirations align with its academic focus
• Transcript of all modules and results from university-level courses taken
• Contact information for two references

Early application is highly encouraged, as the applications will be processed upon reception. Please apply ONLINE formally through the HR system. Applications by Email will not be considered.

Job details

• Contract Type: Fixed Term Contract 36 Month
• Work Hours: Full Time 40.0 Hours per Week
• Location: Limpertsberg Campus
• Internal Title: Doctoral Researcher
• Job Reference: UOL07998

The yearly gross salary for every PhD at the UL is EUR 41976 (full time).