Sara Gettel Email & Phone Number

- Maintained and updated production model to generate public-facing revenue predictions. Oversaw migration of model from DC/OS to Amazon ECS- Developed XGBoost model to recognize incorrectly-denied medical charges on millions of claims. Integrated preliminary model into staging environment. Worked closely with client- facing team to develop experiments, and understand client preferences and risk-tolerance- Developed proof-of-concept model to recognize medical claims that were missing documen- tation. Used tf.keras with categorical embeddings on 1 million claims with highly imbalanced classes. Re-evaluated business proposition of project in response to changing stakeholder priorities- Developed python and groovy templates for automated updates to metrics & alarms in Amazon Cloudwatch, ensuring real-time feedback on model performance and uptime for all deployed models- Identified potential use cases of Amazon SageMaker for distributed machine learning needs and determined it did not meet team’s needs cost-effectively

Watertown, Ma

- Developed scalable tools for analyzing 260 million free-text patient health histories and generated word embeddings with Word2Vec to identify relationships between UMLS clinical concepts within those histories- Developed RNN Keras model to categorize patient health history questionnaires and worked closedly with stakeholder team to understand data from recently acquired com- pany- Mentored two summer interns in developing machine learning models to classify lab results - Developed patient outreach messaging funnel for annual wellness visits and conducted A/B testing, leading to a 2% increase in patient engagement across 500,000 patients- Defined new metrics for clients in Accountable Care Organizations and built monthly re- porting dashboards to highlight client strengths and weaknesses

Greater New York City Area

Created a web app to reduce the risk of auto theft and car break-ins in Boston. Stored and retrieved data on 18,500 crimes with MySQL and processed data with Numpy and Pandas. Developed a density estimation model in Python to predict the least safe places to park in Boston at a given time of day. Deployed an interactive front end using Flask, Bootstrap and AWS.

Cambridge, Ma

Planned and conducted radial velocity observations of dozens of Kepler targets and nearbystars, analyzed observations with Numpy. Measured masses of several planet candidates using non-linear optimization and Bayesian statistical analysis, confirmed existence of new system, to result in publication. Performed simulations of the effect of starspots on radial velocity measurements.

Developed method to calibrate radial velocity measurements in a new wavelength regime, doubling the available bandwidth. Detected new several planets around giant stars, using Monte Carlo simulations. Optimized planet detectability as a function of stellar properties, reducing noise by a factor of two. Adapted IDL optical spectrograph reduction pipeline for current instruments, increasing flexibility of data processing.

Taught introductory astronomy lab, gave planetarium shows and operated small telescopes for student observing projects

Built absorption model of Earth's atmosphere using IDL and the HITRAN database. Took calibration images to aid in the development of an infrared spectrograph.

Developed method to identify 1000 variable stars of type WUMa from a database of 63,000variable stars.

Assisted students with introductory mechanics and E&M classes

Phd, Astronomy & Astrophysics

Activities and Societies: Grad Student Association

Education record

Women in Science bootcamp focusing on shell, GIT, SQL and Python

Master Of Science (Ms), Astronomy And Astrophysics

Activities and Societies: Graduate Student Association

Bs, Physics

Activities and Societies: Society of Physics Students, Sigma Pi Sigma