About me

I am a postdoctoral researcher at the Infosys – Cambridge AI Centre in the Department of Applied Mathematics and Theoretical Physics (DAMTP) at the University of Cambridge. My work focuses on uncertainty-aware machine learning for LHC event generation, developing fast and precise surrogate models for scattering amplitudes that provide accurate predictions with well-calibrated uncertainties.

I am also interested in bridging particle physics and AI, creating methods to extract more precise information from both theoretical predictions and LHC data, building on my experience with global analyses using SMEFT and EDM measurements. Working with AI, I mainly, but not only, focus on the trustworthiness of AI predictions by improving them through appropriate uncertainty introduction and handling, and on the interpretability of those results.

Research interests:

  • Statistical methods in particle physics
  • Surrogate networks
  • Uncertainty estimation with AI
  • Global analyses for SMEFT analyses
  • Amplification of generative networks

News

  • New preprint: Faithful Uncertainties in ML-Based Inverse Problems: GPs, NNs and NTKs for the Triplet PDF, with Sven Krippendorf and Maria Ubiali. Read the paper

  • I gave a talk at BOOST 2026 in Kraków, Poland: When Do ML Uncertainties Fail? - A PDF fit case study. Conference

  • I gave an invited seminar at Nikhef in Amsterdam: From Amplitudes to PDFs: Uncertainty-Aware ML. Talks

Selected publications

All publications
  1. Faithful Uncertainties in ML-Based Inverse Problems: GPs, NNs and NTKs for the Triplet PDF

    Nina Elmer, Sven Krippendorf, Maria Ubiali (2026)

    Preprint · arXiv:2610.30239 [hep-ph]

  2. MadNIS at NLO

    Giovanni De Crescenzo, Javier Mariño Villadamigo, Nina Elmer, Theo Heimel, Tilman Plehn, Ramon Winterhalder, Marco Zaro (2026)

    Preprint · Submitted to SciPost

  3. Forecasting Generative Amplification

    Henning Bahl, Sascha Diefenbacher, Nina Elmer, Tilman Plehn, Jonas Spinner (2026)

    Peer-reviewed · SciPost Physics 20 (2026) 150