Cool stuff with ML and antisense oligonucleotides.

Developed novel fuzzy decision tree model based on kernel density estimation. Designed gradient-based methods using this model’s differentiable loss to train feature transformations for automated feature engineering and dimensionality reduction.

Developed novel classification models and techniques outperforming current SOTA to predict gene interaction from protein sequence data aiding in high-throughput antibiotic discovery.Helped develop and design drug discovery pipeline for polyketide small molecules using genomic and mass spectrometry data.

Redesigned sequence alignment algorithm merging multiple sources of text for a document, achieving improved accuracy, >95% reduction in memory usage, and up to 70% reduction in runtime. Programmed visualizations for analyzing and debugging algorithm performance.Helped implement and design named-entity recognition (NER) model to detect addresses, dates, etc. in documents with >97% precision and recall.

Worked with the Cognitive Neuroscience Laboratory (CNL) at Northwestern University to design and train machine learning models to predict stages of sleep given data from a wearable Fitbit device. Helped build a Fitbit and Android applications as part of a system to allow patients to use these models at home in conjunction with various studies.

Developed game software that stroke victims could play by flexing/relaxing their muscles, in order to help train these people to regain motor function after stroke. Published in Annals of Clinical and Translational Neurology.

Set up and managed cryptocurrency mining computers of varying types, mining Bitcoin and Ethereum. Built system for effective air flow and temperature regulation of these mining computers.

Built a web app for medical practitioners to quickly run predictive medical algorithms assessing the condition iliac vein compression. Published in JACC: Interventions.