CV

Doctoral Researcher

Helmholtz AI, with TUM and ETH Zurich

Machine learning research jointly supervised by Niki Kilbertus (TUM) and Andreas Krause (ETH), funded by Helmholtz AI.

Scholar Representative

Studienstiftung des deutschen Volkes

Elected to represent approximately 1,000 regional fellows, directing a buddy programme, coordinating with the scholarship administration, and organising events including the centenary concert.

Research Assistant

Technical University of Munich (TUM)

Independently devised advanced geometric deep learning methods to improve machine-learning-based interatomic potentials for computational chemistry in Prof. Stephan Günnemann’s group.

Mathematics Tutor

Technical University of Munich (TUM)

Taught two two-hour weekly tutorial sessions for IN0018 Discrete Probability Theory.

Board Member

UniversitätsChor München

Software Engineering Intern

Google

Prototyped generative AI methods for automatic video synthesis in Python.

• Presented the prototype at the internal Beakers & Breakthroughs event to Google Labs Research Director Steve Seitz and Vice President Josh Woodward.

Web Developer

Self-employed

Software Engineering Intern

Google

Developed a novel addressing scheme for Google Cloud Tags in Java.

Student Representative

FSinfo Student Council, University of Passau

• Represented the student body as a full voting member on two professorship appointment committees for AI Engineering and Discrete Mathematics under the Bavarian Hightech Agenda.
• Directed the personal mentoring programme for first-year students three times.
• Led a successful task force to resolve administrative and teaching issues in a laboratory course.

Technology Appointee

Student Choir of Passau

• Developed, administered, and maintained the choir website, server, and mailing list.
• Built a digital tracking system with QR codes for choir rehearsals during the pandemic.
• Edited, mixed, and published recordings of semester closing concerts.

Conference Organiser

University of Passau

Organised and co-moderated the student conference within the 2021 joint meeting of the German and Austrian Mathematical Societies (DMV and ÖMG).

Teaching Assistant

University of Passau

Assisted with undergraduate courses 5200 Algorithms and Data Structures and 5306 Theoretical Computer Science.

On-call Social Volunteer

Bavarian Red Cross (BRK)

Completed errands such as shopping for quarantined and medically high-risk individuals during the first COVID-19 wave, without compensation and with full responsibility for the BRK’s VW T6.

PhD

Technical University of Munich, ETH Zurich, and Helmholtz AI

Machine Learning

• ELLIS joint supervision by Prof. Niki Kilbertus (TUM) and Prof. Andreas Krause (ETH).
• Expected completion in September 2030.

MSc

Technical University of Munich (TUM)

Informatics

• Courses: Machine Learning, Statistical Foundations of Learning, Causality, Reinforcement Learning and Approximate Dynamic Programming, Advanced Deep Learning for Robotics, Computer Vision III, Algorithmic Game Theory, Computational Social Choice, Advanced Algorithms.
• Final grade: 1.0 (best possible); top 2% in cohort.
• Exchange term at UNSW Sydney in the second half of 2024.
• Thesis: Sample-Efficient Reinforcement Learning under Sparse Dynamics.

MSc

University of Oxford

Mathematical Sciences

• Courses: Advanced Statistical Machine Learning, Algorithmic Foundations of Learning, Numerical Linear Algebra, Continuous Optimisation, Theories of Deep Learning, Geometric Deep Learning.
• Final grade: Distinction (best possible); ranked 4th of 45 in cohort.
• Thesis: Flood Forecasting with Graph Neural Networks.

BSc Computer Science, BSc Mathematics

University of Passau

Computer Science and Mathematics

• Final grade: 1.0 (best possible); top 1% across cohorts from 2003 to 2022.

High School Diploma (Abitur)

Apian-Gymnasium Ingolstadt

• Final grade: 1.0 (best possible); ranked 1st of 162 in cohort.

Faculty Award

University of Passau

Award for the best bachelor’s degree in the cohort, with €200 prize money.

Joan Protheroe Scholarship

Lincoln College, University of Oxford

Recognition of musical talent with a £100 honorarium and weekly singing lessons; held in 2022-2023.

DAAD Fellowship

German Academic Exchange Service (DAAD)

Funding of €28,250 for studies at Oxford in 2022-2023, awarded on academic merit.

Studienstiftung Scholarship

Studienstiftung des deutschen Volkes

Scholarship held in 2020-2026, providing a monthly €300 stipend and funding for study abroad in recognition of academic talent and potential contributions to society.

Max Weber Programme

Bavarian State Ministry of Science and the Arts

Scholarship held in 2018-2026, providing a monthly €215 stipend and earmarked grants for talented students in tertiary education.

The Merit of River Network Topology for Neural Flood Forecasting

Proceedings of the 41st International Conference on Machine Learning (ICML 2024)

Nikolas Kirschstein and Yixuan Sun. Proceedings of Machine Learning Research, volume 235, pages 24713-24725, Article 990.

Deep Active Learning for Detection of Mercury's Bow Shock and Magnetopause Crossings

Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2022, Proceedings, Part IV

Sahib Julka, Nikolas Kirschstein, Michael Granitzer, Alexander Lavrukhin, and Ute Amerstorfer. Presented at ECML PKDD 2022; proceedings published by Springer in 2023, Lecture Notes in Computer Science, volume 13716.

Sample-Efficient Reinforcement Learning under Sparse Dynamics

Research in sample-efficient reinforcement learning at the Technical University of Munich.

• Code on GitHub.

A Multi-objective Approach to Safe Reinforcement Learning for Smart Grids

Safe reinforcement learning research at the Technical University of Munich.

• Code on GitHub.

Factory Manipulation with Cooperative Multi-Agent Reinforcement Learning

Reinforcement learning and robotics research at the Technical University of Munich.

• Code on GitHub.

Game-Theoretic Multi-Agent Reinforcement Learning

Essay on game theory and reinforcement learning at UNSW Sydney.

How Transformers Speak: An Interacting Multi-Particle System Perspective

Research on transformers and ordinary differential equations at the Technical University of Munich.

• Code on GitHub.

Flood Forecasting with Graph Neural Networks

MSc thesis on geometric deep learning at the University of Oxford.

• Code on GitHub.

Noise Schedules for Diffusion Probabilistic Models

Research on computer vision and probabilistic modelling at the University of Oxford.

• Code on GitHub.