Avneet Kaur

In this project, we developed a multilingual framework to assess political biases in large language models (LLMs) across different languages, addressing limitations in understanding language-sensitive retrieval of political opinions. The approach involves data collection, model training, bias measurement, and the development of a multilingual framework. The anticipated outcomes include insights into language-sensitive opinion retrieval and the establishment of a foundation for more inclusive and language-aware LLMs.
Featured as the most innovative project in Cohere for AI Blog post

In this project, I contributed to the development of Our Multilingual Climate Chatbot project that aims to make climate research and education accessible to all.
Code / Featured as the winner project in Cohere for AI Blog post

Danish parcel delineation is an important task that currently has to be done manually. In this work, we automate this laborious and error-prone process. Based on real data from Danish agriculture, we applied deep learning methods to efficiently detect parcels from freely available satellite images. To that end, we created a complete data pipeline, from data collection over to processing, then building and evaluating the model.
Code / PDF / Slides

We developed an open, robust and automated preprocessing pipeline for PET data using existing state-of-the-art neuroimaging software and implemented using Nipype. To test the pipeline for robustness and computational reproducibility, the pipeline was tested on three datasets from different scanners and tracers across different computational environments.
Code / PDF / Slides