Research Projects & Publications

A real-time activity recognition pipeline leveraging machine learning to classify motion data and estimate MET (Metabolic Equivalent of Task) values. Built with a React Native front-end and a Python backend, the system streams accelerometer data, preprocesses it in sliding windows, and posts it to a trained model endpoint for inference. Results are displayed live, enabling continuous energy expenditure monitoring.

CPR-AI is a machine learning-based system for recommending clinical procedures based on patient demographics and conditions. The system uses a Random Forest model trained on clinical data to predict the most appropriate next procedure for a patient.

SRSBPI is a novel brain prognostic index for lung cancer patients with brain metastases, aiding in therapeutic strategies. Developed for the Department of Oncology-Pathology, Karolinska Institutet/Thoracic Oncology Center, Karolinska University Hospital.

The BPNN is a neurobehavioural analysis deep learning tool developed at the Meletis Lab, Karolinska Institutet, as part of my Master Thesis titled 'Direct Behaviour Prediction from Miniscope Calcium Imaging Using Convolutional Neural Networks.'