I develop data products to solve people challenges. I lead initiatives that:Automatically categorize employee survey responses which significantly scaled the number of reports 10x, enhanced accuracy and eliminated a costly software contract. Implemented a comprehensive organizational health metric, integrating multiple univariate metrics into a singular north-star metric to gauge team health effectively.

I led and implemented multiple projects to personalize our user's experiences along with various statistical analyses. These have included:Improving organizational leader experience to solve that organizational leaders desired higher level content, but we didn't know who they were within our populations. After collecting and processing the relevant data, we designed and implemented a random forest model to automatically identify leaders at each organization.Personalizing of nudges to individual users. Using a user's historic engagement and their role, we create a recommendation system to suggest the best nudge for each user.

Hire, structure, and lead a team of 6 data scientists and data engineers to automate the medical device's procedures and diagnostic technique. I am responsible for the AI/machine learning models and their continuous training infrastructure (MLOps).I constructed an NLP pipeline to extract clinical diagnoses after redacting PHI/PII from hospital PDFs.

At MDS, I led data science projects to create new data-focused products across multiple industries. This work has produced: 2 CPG focused data products that enable brands to understand promotional behavior at individual convenience stores and inform them of demographic differences between their consumers.An NLP product to understand medical literature publications. This product provides researchers with a tool to summarize the vast quantity of literature surrounding their question of interest. Utilized medical device readings to categorize tumor characteristics. Consumer behavior profiles to target high-spending individuals in consumer-tech.From IoT manufacturing sensors, model likelihood of faults and errors to reduce the product's impurities. Data ingestion pipeline to automate the standardization and QA of multiple real-estate data sources.

Led consolidation of manual adjudication codes. Interviewed all stakeholders to determine their needs and determined path forward. I highlighted most valuable automation targets.

Fraud department sought improvement in dollar return on Modifier 25 investigations; they wanted specific claims to scrutinize. I constructed and productionized a PySpark random forest model that output likelihood of fraud, waste or abuse for most recent 3 months of claims; investigators in midst of pursuing first batch of claims when I relocated.With a list of illegal Special Enrollment Periods enrollees, I visualized, explored and tested differences between known illegal SEP enrollments and general SEP enrollments. I referred 40 high-cost members for investigation and uncovered manipulation of out-of-state laboratory benefits that resulted in a product revamp and expected savings of $10 million in 4th quarter 2016.Given limited Medicare marketing budget, the marketing team hoped to optimize spending towards members unlikely to re-enroll. I constructed predictive models for each Medicare product that output a member's likelihood to churn. This improved identification of churners from 41% to 63% and attrition dropped from 7% to below 5%.Authorization-based referrals to care management were set to expire; this required a new system to maintain the referral source. I modeled each member's future medical cost to prioritize referrals. The new method included a member's medical history, demographic information, and all authorization types; this referred additional $300 million in medical cost to care management over old referrals.Marketing department dissatisfied with direct-mail targeted by zip code, they hoped to use individual household information. I clustered prospects with K-means, described the story of each cluster's characteristics for the creation of relevant copy. We saw the expected product purchased 10% closer to plan than prior year.

Taught intro to business statistics to undergraduate students. Topics covered include: distributions, hypothesis testing, and linear regression.

Helped construct book on text mining and content analysis (a more structured approach to text mining). Examined NASA management response to the Apollo I, Challenger, and Columbia disasters.Logistic regression model for NFL player arrests using player tweets.