1) predicting the next migraine attack of a given patient (with 24 hours lead-time) given its health indicators from smart devices, and 2) identifying some Covid-19 indicators of the user such as body temperature from the facial video record, and heart rate variability (HRV) from user's finger recording on the camera.

Learning neural embeddings of policy texts, to further identify which types of the data of a user is shared/stored by a company.

Processing noisy ECG images to construct digital ECG signals.

9 Month Thesis on ICU (intensive care units) Mortality Prediction of Patients: I implemented a deep Markov model to learn the latent phases of the ICU patient, then built an attention network on top of the latent variables to predict the mortality risk of the patient. On a well-established benchmark health dataset MIMIC-3, I have experimentally shown my algorithm beats all the state-of-the-art ICU mortality predictors. Thereby, my thesis has been regarded as an outstanding achievement, and received the grade 6.0/6.0 . Besides, I submitted a paper to KDD 2021 Conference, which we expect to be accepted by May 2021. You can find the free arXiv version here: https://arxiv.org/pdf/2102.04702.pdf .