Lukas Aichberger

About

👉 aichberger.github.io

Experience

• Doctoral Candidate in Artificial Intelligence at ELLIS - European Laboratory for Learning and Intelligent Systems
  Oct 2022 - Present · 3 yrs 9 mos

  • Co-supervised by Prof. Sepp Hochreiter at JKU and Prof. Yarin Gal at the University of Oxford. • Published 14 papers with over 230 citations to date, including 4 first-author papers at NeurIPS and ICLR.

• Johannes Kepler Universität Linz (4 yrs 10 mos)
  • Doctoral Researcher
    Oct 2022 - Present · 3 yrs 9 mos

    • Worked with Prof. Sepp Hochreiter and his Institute for Machine Learning (IML) on Uncertainty in Deep Learning. • Proposed G-NLL, SDLG, and QUAM: state-of-the-art methods to quantify uncertainty of classifiers and LLMs.

  • Pre-Doctoral Researcher
    Sep 2021 - Sep 2022 · 1 yr 1 mo

    • Designed, implemented, and evaluated models for compound-protein activity prediction from molecular representations for drug discovery. • Engineered an HPC showcase cluster and implemented distributed ML algorithms for the EU-funded project ELISE (Grant agreement ID: 951847)

• Independent ML Researcher & Engineer at Aichberger Solutions
  Mar 2023 - Present · 3 yrs 4 mos

  • Developing end-to-end ML prototypes, from data pipelines to model training and deployment.

• Machine Learning Research Intern at Apple
  May 2025 - Sep 2025 · 5 mos

  • Developed an uncertainty-guided search strategy for diversifying LLM reasoning, doubling valid solution rates. • Post-trained LLMs on diverse reasoning traces, outperforming decoding baselines on hard math benchmarks.

• Doctoral Researcher at University of Oxford
  Oct 2024 - Apr 2025 · 7 mos

  • Worked with Prof. Yarin Gal and his Oxford Applied and Theoretical Machine Learning Group (OATML) on Uncertainty Quantification for LLM Function-Calling • Collaborated with Prof. Philip Torr and his Torr Vision Group (TVG) on Adversarial Robustness of Multimodal OS Agents