Research & Projects |

Intraday Electricity Prices and Optimal Liquidation

Thu, 01 Jan 2026 00:00:00 +0000

How should an electricity position be liquidated when trading activity and liquidity change sharply before delivery? At Aarhus University, in collaboration with Aros Commodities, I developed stochastic models combining self-exciting jump arrivals, time-to-delivery effects, and empirically estimated jump sizes. I examined auction-related price drops and recovery and represented these effects through scheduled stochastic transitions. A memory-efficient Python framework supported maximum-likelihood estimation and out-of-sample evaluation on sparse, one-second quote data. I also formulated Hamilton–Jacobi–Bellman equations for liquidating long positions under price uncertainty and execution costs. The ongoing research connects interpretable price models with execution decisions.

Probabilistic Trajectory Prediction

Mon, 31 Oct 2022 00:00:00 +0000

Automated driving requires predictions of where other road users may move and how uncertain those predictions are. At Bosch, I developed a Gaussian regression-based method for trajectory prediction with explicit uncertainty estimates. I evaluated forecasting accuracy and uncertainty calibration against production baselines and translated Python research prototypes into C++ systems with defined reliability requirements. The broader programme included analysis of more than 20,000 hours of driving data and technical leadership of an eleven-person environment-prediction team. The patent application and demonstration document the trajectory-prediction method, connecting probabilistic modeling with the practical constraints of real-time software.

Stochastic Models of Higher-Order Networks

Mon, 15 Dec 2025 00:00:00 +0000

Many networks involve interactions among groups, rather than only pairs of entities. During my PhD at Aarhus University, I developed spatial and dynamic models of these higher-order relationships with collaborators. We established limit theorems for network statistics, including degree distributions, edge counts, and topological characteristics. I implemented high-performance Monte Carlo algorithms in Python and C++ to investigate finite-sample behavior and assess theoretical predictions. Fitting models and conducting hypothesis tests on scientific collaboration networks connected the mathematical results to empirical data. The research is documented in my thesis, work on higher-order network topology, and dynamic hypergraphs, alongside public simulation software and a collaboration dataset.