Investigating whether image inversion improves detection of
AI-generated images. Adapting techniques from image editing,
my method learns from the original image, its inverted noise
and its reconstruction.
Finding five-player lineups under budget, team and tournament
constraints. Built in Python, with a Streamlit interface,
packaged with Docker and deployed on Azure.
03 / RecommendationRecommender systems · Group project
Your next favourite?
Beer recommender system
Helping people discover beers through patterns in user ratings.
We compared collaborative filtering methods and explored a hybrid
approach that balances personalised predictions with broader
recommendation coverage.
SVD / hybrid recommendationYour rating history → popular picks / personalised picks
The research approach
11–45 reviewsPopular picks
Community favourites
More than 45 reviewsPersonalised picks
Patterns learned with SVD
Discover your next beer
Behind the workBergen / Norway
Kjell Runar Horvli / Bergen, NO
Curiosity, put to work.
About me
I’m a master’s student in Information Science at the University
of Bergen, focusing on machine learning and AI. My thesis explores
detecting AI-generated images through image inversion.
Alongside my studies, I’ve contributed to Norwegian-language
LLM safety research through SafeLLM and worked as a teaching
assistant in machine learning and AI courses.
