-implementing and extending the mask2former model into our detectron2 based vision pipeline for enhanced predictions on challenging objects-experimenting with other options to enhance predictions(adding features from our 3d data, pretraining the resnet backbone with VICReg on our data...)-various other smaller data science tasks regarding the optimization of our vision pipeline

Worked at the coroprate client division. Mainly involved in the preparation stage for a predictive model that would identify customers that would need financing in the upcoming time so that the clients could be targeted soon enough to ensure they receive the financing from us.