Private consultant for machine learning (ML) research with a focus on medical/health applications. Currently assisting researchers at SUNY Upstate Medical Center with designing and building ML models to better understand Alzheimer's etiology via genome sequence data.

Goal: To improve continuous, in-the-wild Parkinson’s Disease (PD) symptom detection from wearable sensors using machine learning (ML) algorithms.Explored several domain adaptation algorithms, which leverage unlabeled data from the user to improve “generic” models trained on data collected from a training population, and applied them to PD tremor detection. Developed a novel, “stratified” technique for weakly supervised learning, which aligns better with the application of PD tremor detection than standard weakly supervised techniques and enables training on data collected in the wild (see 2017 EMBC paper). Conducted experiments evaluating whether performance of ML algorithms on laboratory data generalizes to the wild: findings indicate that products expected to operate in wild settings should be trained on wild data.Advisors: Fernando De la Torre and Jessica Hodgins

Goal: To develop computer vision algorithms for automatic segmentation and analysis of high-resolution ultrasound images.Imaged nerve injuries in rats with high-resolution ultrasound to develop algorithms for automated detection of nerve regeneration. Collaborated with plastic surgeon Dr. Vijay Gorantla from the University of Pittsburgh Medical Center to validate findings on hand transplant patients.Advisor: George Stetten

Goal: To understand and predict surgeon behaviors during surgical tasks.Worked with metrics computed from kinematic data output by the da Vinci Surgical System. Implemented several algorithms and developed modifications for personalization to predict metrics on future surgical tasks given metrics on earlier surgical tasks.Manager: Anthony Jarc

Goal: To improve control of the Anatomically Correct Testbed (ACT) Robotic hand for expressive piano performance.Designed an experiment to help quantify musical versus non-musical piano performance. Submittedan application to the Institutional Review Board for human subjects approval. Analyzed data usingMATLAB. Improved control strategies for the ACT hand, allowing it to perform piano musically.Advisor: Yoky Matsuoka

Tutor university students in pre-calculus, calculus (single, multivariable and differential equations), and matrix algebra courses. Prepare exam review sessions.

Assist with autopsies, specimen preservation and tissue discard. Prepare autopsied patients for release to funeral home. General setup, clean up, and stocking.