Bachelor’s thesis, University of Agder; in collaboration with Red Rock AS · 2025
A prototype that uses YOLOv11 to detect components in hydraulic schematics, lets users review the detections, and generates Structured Text from their connections.
University of Agder · 2025
A local evaluation pipeline comparing ROUGE, BERTScore, and an LLM judge on scientific summaries from three models. The methods showed weak agreement on individual summaries.
University of Agder · 2026
A Chinese Checkers agent trained with PPO and action masking in a custom PettingZoo environment, with analysis of training and tournament play.
An interactive Python simulation with adjustable road capacity and demand. It uses car-following and lane-changing models to illustrate how congestion can return after roads are widened.
Python · Tkinter · IDM · MOBIL
Norwegian Coastal Administration
Returned to develop AISyRisk, the Coastal Administration’s grounding-risk calculation system. Further developed AIS-based data pipelines and models, and improved internal risk-calculation services.
Norwegian Coastal Administration
Developed a machine learning model to analyse and predict vessel sailing patterns from historical AIS data. Built a data pipeline for model training, applied geospatial analysis, and presented findings to the team and management.
Align Racing UiA
Developed and tested LiDAR-based autonomous driving systems. Implemented path planning and obstacle detection with ROS and Python.
Technical University of Munich
Data analysis, artificial intelligence, and machine learning.
University of Agder
Grimstad, Norway
University of Agder
Thesis: AI-Driven Engineering Automation: From Technical Drawings to Structured Code. Used YOLOv11 for component recognition in collaboration with Red Rock AS.