Time series foundation models · synthetic data generation · zero-shot probabilistic forecasting · multivariate forecasting with covariates · recurrent architectures · constant-cost streaming inference
Oct 2025 – now
PhD in Artificial Intelligence
Johannes Kepler University Linz · Institute for Machine Learning
Advisor: Univ.-Prof. Dr. Sepp Hochreiter. Topic: Time Series.
Mar 2023 – Aug 2025
MSc Artificial Intelligence
Johannes Kepler University Linz
Thesis: Parameter-Efficient & Memory-Efficient Fine-Tuning of Convolutional Neural Networks for Image Classification, advisor Univ.-Prof. Dr. Günter Klambauer. Theoretical and empirical analysis of LoRA, VeRA, GaLore and normalization-based fine-tuning.
Mar 2023 – now
MSc Legal and Business Aspects in Technics
Johannes Kepler University Linz
Oct 2020 – Mar 2023
BSc Technical Physics
Johannes Kepler University Linz
Completed in 5 semesters (standard duration: 6). Thesis: Tuning of Purcell enhancement of GaAs quantum dots in circular Bragg gratings by wet chemical etching, advisor Univ.-Prof. Dr. Armando Rastelli. Optics lab measurements of entangled photons from GaAs quantum dots.
Feb 2017 – Sep 2019
Apprenticeship with university entrance qualification (Matura)
HTL Leonding, BFI OÖ, Volkshochschule Linz
2026
TiRex-2: Generalizing TiRex to Multivariate Data and Streaming NeurIPS 2026Oral
Patrick Podest*,
Marco Pichler*, Elias Bürger, Levente Zólyomi, Bernhard Voggenberger, Wilhelm Berghammer, Daniel Klotz, Sebastian Böck, Günter Klambauer, Sepp Hochreiter.
arXiv:2607.01204* equal contribution
Jul 2026 – now
PhD Research Scientist Intern
NXAI GmbH · Linz, Austria
Seven-month internship.
Oct 2025 – now
Research Associate (third-party funded)
Institute for Machine Learning, Johannes Kepler University · Linz, Austria
On leave since Jul 2026 for the NXAI internship. Joint first author of TiRex-2, a recurrent xLSTM-based multivariate time series foundation model with past and future covariates.
- Bidirectional time mixer and asymmetric grouped-attention variate mixer: future-known covariates are read in both directions while targets stay strictly causal
- Synthetic coupling pipeline that composes diverse multivariate training samples on the fly from univariate corpora
- Constant-cost streaming inference, stable far beyond the training context; zero-shot evaluation on GIFT-Eval and fev-bench
Aug 2021
Intern, Infrastructure & Service
LIWEST Kabelmedien GmbH · Linz, Austria
Fault analysis and network integration for HFC, FTTH and 5G FWA customer connections.
Jul 2020 – Sep 2020
Electronics Technician, Handover Automation
KEBA Group AG · Linz, Austria
Assembly, soldering and cabling of automation components for logistics systems.
Oct 2019 – Jun 2020
Civil Service
Kindergarten & senior citizens' home · Upper Austria
Sep 2015 – Sep 2019
Apprenticeship as Electronics Technician
KEBA Group AG · Linz, Austria
Applied electronics (main module) and network engineering (special module). Final apprenticeship examination passed with honors.
2026
Oral presentation, NeurIPS 2026 · TiRex-2
One of 112 orals among 30,709 main-track submissions (top 0.4%).
since Sep 2020
Study grant for self-supporting students · Austrian Study Grant Authority
Sep 2024 – Aug 2025
Curricula Examination Committee, student member
Student Union, Johannes Kepler University Linz
Reviewed curricula for legal compliance, academic feasibility and structural consistency.