I'm a software engineer at Pine59 in Oslo, building location intelligence that helps businesses understand how places and populations change. Before that, I did my PhD in deepfake detection at Colourlab, Norwegian University of Science and Technology (NTNU).
My research looked at how detectors generalize to unseen manipulations, whether they are fair across demographic groups, and whether benchmarks really measure what they claim to. I was advised by Kiran Raja, Marius Pedersen and Luisa Verdoliva.
I build the data pipelines behind Pine59's mobility and location products. Most of my work turns aggregated, anonymized telco signals into products people can use: origin–destination matrices and traffic data, visitor and trade-area analytics on H3 hexagon grids, and the road, rail and transit map layers that tie them to real places. Pine59 grew out of Unacast's location insights business in 2026.
Researched deepfake detection that holds up outside the lab: models that generalize to unseen manipulation methods, explain their decisions, and treat demographic groups fairly. Thesis: Generalising Deepfake Detection Towards Robust, Interpretable and Fair Models.
We annotated five popular deepfake datasets with 41 demographic and non-demographic attributes, about 65 million labels in total, and used them to measure bias in state-of-the-art detectors.
