cv

ELLIS PhD Student

Institute for Machine Learning, Johannes Kepler University Linz, Austria

Currently building Large Language Models. Previously worked on Few-Shot Learning, Meta-Learning and Domain Adaptation.

• Large Language Models
• Machine Learning
• Deep Learning
• Natural Language Processing
• PyTorch/CUDA

Master Thesis Student & Research Assistant

FZI Forschungszentrum Informatik, Karlsruhe, Germany

Master Thesis in cooperation with Institute of Measurement and Control Systems: Motion Planning for Automated Vehicles in Uncertain Environments.

• Application of Partially Observable Markov Decision Processes (POMDPs) in the domain of automated driving.
• Development of a Monte-Carlo Tree Search based POMDP planner for interactive urban intersection scenarios in C++.

Student Research Assistant

FZI Forschungszentrum Informatik, Karlsruhe, Germany

Contributed a package for visibility computations to an internal simulation environment for automated driving.

• Software development for computational geometry applications in C++ and Python.
• Development of packages for visibility computations using CGAL for a simulation environment in the field of automated driving.

Student Research Assistant

Karlsruhe Institute of Technology, wbk Institute of Production Science, Germany

Worked on a project to restore a machining center of the type DMC 60H.

• PLC programming for a machining center of the type DMC 60H with Siemens SIMATIC Step 7
• Conception and design of mechanical components and a test rig for visual measuring of position accuracy in industrial robots with Siemens NX

PhD Artificial Intelligence

Johannes Kepler University Linz

Artificial Intelligence

2 Semester Exchange Program

San Jose State University

Computer Engineering

• Important Courses: Embedded Control System Design, Real-Time Embedded System Co-Design, Operating Systems, Advanced C++ Programming

Master of Science

Karlsruhe Institute of Technology

Mechatronics and Information Technology

• Master Thesis: Motion Planning for Automated Vehicles in Uncertain Environments
• Important Courses: Embedded Control System Design, Real-Time Embedded System Co-Design, Operating Systems, Advanced C++ Programming

Bachelor of Science

Karlsruhe Institute of Technology

Mechatronics and Information Technology

• Bachelor Thesis: Machine Control and Industry 4.0 Functions for a Machining Center of Type DMC 60H

Addressing parameter choice issues in unsupervised domain adaptation by aggregation

International Conference on Learning Representations

We introduce an aggregation method that tackles the problem of choosing algorithm hyper-parameters in unsupervised domain adaptation. We compute several models using different hyper-parameters, and propose a method for a linear aggregation of the models, that extends weighted least squares to vector-valued functions, e.g., deep neural networks.

Few-shot learning by dimensionality reduction in gradient space

Conference on Lifelong Learning Agents

We introduce SubGD, a novel few-shot learning method which is based on the recent finding that stochastic gradient descent updates tend to live in a low-dimensional parameter subspace.