Alex Podgorny

About

Since September 2026, I have been a postdoctoral researcher at Centre Borelli (ENS Paris-Saclay), working with Anne Sabourin, after a first postdoctoral position at ENSAI and IRMAR (Rennes 2 University) within the ANR project FunMathStat. My research lies at the intersection of extreme value theory, statistical learning, time series, and kernel methods. I currently work on learning conditional quantiles in extreme regimes for time series, as well as on dimension reduction for conditional tail inference and invariant RKHS methods for functional data analysis.

My CV is available here (PDF).

Research

Current postdoctoral project

Since September 2026, I have been working at Centre Borelli with Anne Sabourin on a problem at the intersection of extreme value theory, time series, and statistical learning. The goal is to estimate a future quantile of a time series from its recent dynamics, conditionally on the occurrence of an extreme event. The project relies on the spectral tail process of regularly varying time series and on a quantile-regression procedure built from the most extreme observations. One of the main theoretical objectives is to derive non-asymptotic excess-risk bounds that account for both the complexity of the predictor class and temporal dependence. This work is ongoing.

First postdoctoral position

From October 2025 to 2026, I held a first postdoctoral position at ENSAI and IRMAR within the ANR project FunMathStat. My work focused on nonparametric tests invariant under group actions in reproducing kernel Hilbert spaces (RKHS), with applications to functional data. This project led to the submitted paper listed below.

Summary of my PhD

My PhD research was carried out in extreme value statistics, with a particular focus on the study and estimation of extreme quantiles of high-dimensional conditional distributions. It was structured around three main axes: (1) the construction of confidence intervals for extreme quantiles in general domains of attraction, (2) the development of an original dimension-reduction method for estimating the conditional tail index, and (3) a comparative study of modern approaches (machine learning, dimension reduction) for conditional extreme quantile estimation.

Published papers

• Gardes, L., Maistre, S., and Podgorny, A. (2025). Asymptotic confidence intervals for extreme quantiles in a maximum domain of attraction, Electronic Journal of Statistics, 19(1), 1103–1132. [DOI] [HAL]
• Gardes, L., and Podgorny, A. (2025). Dimension reduction for the estimation of the conditional tail-index, Scandinavian Journal of Statistics, 52(3), 1444–1476. [DOI] [HAL]
• Gardes, L., and Podgorny, A. (2025). Extreme Conditional Quantile Estimation in High Dimensions: A Comparative Study, International Statistical Review, early view. [DOI] [HAL]

Submitted papers

• Giacofci, M., Meynaoui, A., and Podgorny, A. (2026). A Kernel Two-Sample Test Invariant under Group Action with Applications to Functional Data, preprint. [HAL]

Talks

• A. Podgorny. Asymptotic confidence interval for extreme quantiles, 13th International Conference on Extreme Value Analysis (EVA 2023), University of Milan, May 2023.
• A. Podgorny. Dimension reduction for conditional tail-index estimation, 55th “Journées de Statistique” (JdS 2024), University of Bordeaux, May 2024.
• A. Podgorny. Two-sample MMD test invariant under group action, 57th “Journées de Statistique” (JdS), Université Clermont Auvergne, June 2026.

Teaching

I have completed 204 hours of teaching for students with varied academic backgrounds:

• University of Strasbourg: 192 hours of biostatistics for biology undergraduates — 144h in L2 (2022–2025) and 48h in L3 (2022–2024).
• MVA Master's program, ENS Paris-Saclay: 12 hours of functional-analysis refresher teaching in September 2026 (6h lectures and 6h tutorials), covering in particular L^p spaces, Hilbert spaces, distributions, and Fourier analysis.

Scheduled teaching: in November 2026, I will be responsible for the Extreme Values course at ENSAI (18h: 6h lectures and 12h R practical sessions). I have already designed the full set of lectures, practical sessions, and the final assessment.