Joined Evolent in 2024 when they acquired the Machinify Auth product I helped build for $32 million.

* Served as the lead data scientist on the Machinify startup team that built software acquired for $32 million:https://www.marketscreener.com/quote/stock/EVOLENT-HEALTH-INC-22363225/news/Evolent-Health-Inc-agreed-to-acquire-Certain-assets-of-Machinify-Inc-for-32-million-46997864/* Fine-tuned and used large language models (LLMs) to speed up prior authorization approvals (we never use AI to deny care). Doubled the rate of automatic approvals for prior auth requests, resulting in timely care for more patients.

* Built machine learning models to predict imminent suicide attempts and imminent drug overdoses to perform targeted interventions and save lives. - Wrote complex SQL queries to build dataset of 346 features derived from medical and pharmacy claims data, using rolling windows for each patient to take advantage of all historical data for model training. - Trained the models using PySpark in an AWS EMR cluster to maximize performance on large dataset.* Identified gaps in care for a prospective new customer based on their medical and pharmacy claims data using Python and SQL, which resulted in a multi-million dollar sale and partnership with the customer.

* Founded the Medicare Survey Data Science team. Used machine learning to improve our patient satisfaction scores on the CAHPS survey, which determines our Medicare Star Ratings, federal bonus revenue, and new membership volume.* Led the offshore Ireland data science team. Taught them how to build and deploy performant ML models related to Medicare Star Ratings.* Built two machine learning models to predict members’ responses to the CAHPS survey. One model predicts whether members will report experiencing delays in medical testing and the other predicts whether their doctor will review their full set of prescriptions with them, two key metrics in the federal quality ratings. Worked with the business to design targeted intervention and outreach programs that are projected to save over $1 million for the company via federal bonuses for improved Medicare Star Ratings based on improved patient experience survey results.* Deployed the models using PySpark into a Hive production environment and make monthly predictions for each of our 7 million members, which the company uses to perform targeted interventions.* Connected data from an internal survey of our members with thousands of features gleaned from numerous member-level data sources (claims, diagnoses, disabilities, pharmacy, finance, member engagement, utilization, provider, geographic, demographic, and more) to train models.* Built 21 explanatory models to find actionable insights and identify populations for outreach for all CAHPS survey questions to improve patient satisfaction.

* Founded the Medicaid Data Science team and mentored junior members of the team.* Built a machine learning model to predict which pregnant women will have low birth weight babies, resulting in more effective prenatal care interventions, healthier babies, and decreased NICU costs.* Built a machine learning model to predict mortality in the elderly/disabled population and offer to help terminally ill patients navigate the complex process of enrolling in hospice if the patient and their doctor both believe that is their best option for a more comfortable and dignified death. In addition to these advantages, hospice also reduces medical expenses, and this program saved $650,000 for the Medicaid plan in the first year.* Built a machine learning model to predict which members will be hospitalized, resulting in improved preventative care, decreased hospitalization rates, and decreased inpatient costs.* Created an extensive data map to describe each member using features relating to health, demographics, and finance, for use in machine learning datasets and business intelligence reports.

SCRB 187: Neuroscience and Ethical Philosophy. I won a teaching award for this course (see Awards section).

Summer research project studying microtubule biophysics in Prof. Marileen Dogterom's lab