My interests are at the interface of digital health and machine learning. I currently work on applications of machine learning in healthcare at Philips Research with a special interest in modelling biosignals, sequence similarity search, and incorporating prior knowledge into AI models. Previously, I obtained my PhD in neural engineering and was a data scientist at Stitch Fix. Author of Machine Learning for Digital Health
My work at Philips centers on developing clinical analytics for patient monitors using machine learning to enable personalized care, empower care givers with AI-powered reasoning tools, and improve patient & operational outcomes.
I build realtime clinical risk indicators that are deployed on patient monitors and central stations.
Patient similarity aims to identify cohorts based on a set of characteristics (like demographics, vitals, labs, medical history, and treatments) to enable applications like case-based comparisons for clinical decision support and to compare treatments across similar cohorts.
A key requirement of clinical models is that they have to be explainable so that the nurse or clinician can understand the risk factors. The models should also be editable so we can identify when then algorithm makes an error and fix it before deployment.
I develop deep learning algorithms using temporal convolutional neural networks, recurrent neural networks, and transformers for physiological signals collected from wearable devices, including electrocardiograms (ECG), photoplethysmography (PPG), phonocardiograms (PCG, heart sounds), and arterial blood pressure (ABP). Many of these solutions have come in first place at the PhysioNet Challenges. See publications below.
See all publications on Google Scholar