Research & Projects

This page gives an overview of my research projects and related work, grouped by theme.

Mechanistic interpretability & AI

Mechanistic interpretability of variable assignment in a Transformer-based code model

One-month team research project, MSc Logic, University of Amsterdam (UvA), 2026

As a small team, we investigated how a Transformer-based language model represents variable assignment in Python code. Using a synthetic dataset of assignment chains with identifier reuse, we compared immutable values and mutable lists with aliasing behaviour. We performed causal interventions on the residual stream to probe how the model tracks intermediate values and how performance depends on chain length.

  • Constructed a synthetic Python dataset isolating identifier reuse across assignment chains.
  • Compared the model’s behaviour on immutable values versus mutable lists and aliasing.
  • Used causal interventions on the residual stream to study which intermediate values the model attends to.
  • Observed that the model is accurate on simple cases but performance degrades for longer chains, with no clear, robust heuristic for value tracking emerging.

This team project grew out of the Machine Learning and Language Models and Deep Learning courses in my MSc Logic programme.

Quantum computing & quantum information

Variational quantum algorithm for the vibrational spectrum of SO₂

Group course project, Near-Term Quantum Computing, UvA, 2025

In the Near-Term Quantum Computing course, our team implemented a variational quantum algorithm (VQA) in Qiskit to probe the vibrational spectrum of the SO₂ molecule on a simulator. The project focused on near-term, hybrid quantum–classical algorithms for small molecular systems.

  • Designed a VQA ansatz and cost function suitable for vibrational spectra on near-term devices.
  • Implemented the algorithm in Qiskit using a classical–quantum optimisation loop on a simulator.
  • Analysed the limitations of simulation-only implementations and discussed requirements for running such algorithms on noisy hardware.

For me, this team project connects my technical-physics background with my interest in quantum information and near-term algorithms.

Theses

VHDL test environment for a soft microcontroller on FPGA

Bachelor’s thesis, BSc Technical Physics, Vienna University of Technology (TU Wien), 2021

For my technical physics thesis, I built a VHDL-based test environment on an FPGA around a soft microcontroller.

  • Implemented the test environment in VHDL on an FPGA platform.
  • Configured the soft microcontroller and created C test scenarios to validate its behaviour.
  • Programmed an interface between the microcontroller and a Siemens component to test system-level integration.

This work was closely connected to my internship at Siemens Austria, where I worked on FPGA-based systems and microcontroller integration.

Category-theoretic modelling and priming in linguistics

Bachelor’s theses, BA Linguistics, University of Vienna, 2025

In my linguistics degree, I completed two theses – one in formal semantics and one in psycholinguistics.

  • Formal semantics / category theory: modelled selected linguistic phenomena using functors and monads, exploring category-theoretic approaches to meaning composition.
  • Psycholinguistics / Primantik: designed and ran a priming study on associative and semantic word relations, analysing differences in priming between associative links and purely semantic relations.

These theses connect my interests in logic, semantics, and language with empirical methods and formal tools.

Applied research & engineering internships

Austrian Institute of Technology (AIT), Vienna

Intern, 2024, 2023, 2019

Across several internships at the Austrian Institute of Technology, I worked on environmental data acquisition, front-end and back-end development, and sensor network demonstrators.

  • Developed an Angular frontend and automated end-to-end tests with Cypress; implemented a Java backend with a REST-compliant API and relational-database interface.
  • Programmed applications for an environmental data acquisition project, including an interactive TypeScript/Leaflet map and data-validation components in Java/Proj4J.
  • Built a test setup to demonstrate air-quality measurements of the UWEDAT sensor network solution and developed Python scripts for visualisation and sensor control.

Siemens Austria, Vienna

Intern, 2020

  • Implemented a VHDL-based test environment on an FPGA around a soft microcontroller.
  • Configured the microcontroller, wrote C test scenarios, and programmed an interface to a Siemens component to test integration.

Institute of Applied Physics, TU Wien

Intern, 2018

  • Motorised the X–Y manipulators of a vacuum chamber using stepper motors to enable precise, remote positioning.
  • Designed mounts in CAD and integrated the motors with the existing control system.

These internships gave me practical experience bridging theoretical ideas with concrete implementations in software, hardware, and experimental setups.

Schools, conferences & scholarships

I have complemented my formal degrees with summer schools and conferences in logic, linguistics, and computer science, including ESSLLI 2025 (Bochum) and Eastern Generative Grammar 2024 (Braşov), where I followed courses on language model programming, neurosymbolic learning, temporal logics, and formal semantics.

My studies have been supported by several merit scholarships from the University of Vienna, an Erasmus+ study‑mobility grant, and a conference scholarship for ESSLLI 2025.