Data Scientist
Ex Data Scientist at Claritev · New York, USA
I build production ML for healthcare claims and population health. At Claritev, I worked on pricing analytics across 2M+ monthly claims and calibrated risk and disease-rate models for 1M+ members.
Healthcare document search, trading analytics, and ML execution systems.
I built a drug-document search app that answers questions with passages from Health Canada and Ontario sources.
Reported evaluation: 87.5% Recall@3 and 100% citation validity.
I built a research assistant that answers backtest questions through 21 read-only tools over DuckDB.
Saved evaluation: 23/50 tasks passed (46%). Tool execution: 100%. Aug 30, 2026 run.
I built a leakage-safe research and execution platform for trade filtering and position sizing across five futures markets.
Eight model families across five markets. Walk-forward, out-of-sample, and Monte Carlo validation.
I present these as research and analytics applications. My trading workbench is not an investment advisory product and does not predict future returns.
I am a data scientist with 6+ years of experience across healthcare, financial services, and analytics. My healthcare work includes claims repricing, calibrated risk scoring, disease-rate prediction, and production model monitoring. I use Python, SQL, R, and Snowflake, and work with clinical and engineering teams to turn model outputs into population segments and care-management priorities.
Aug 2023 to Jan 2026
Claritev
Streamlined ProPricer analytics across 2M+ monthly claims and developed calibrated risk-scoring and disease-rate models for 1M+ members. Tuned claim-risk thresholds to lift identified savings about 12%, and supported production pipelines tied to up to $22M in potential annual savings per client. Automated Python and SQL workflows, operated the Snowflake-based CERES XGBoost platform, and applied CI/CD, Docker, and model monitoring. Partnered with clinical and engineering teams on population segments and validation documentation, reducing QA review cycles 40%.
Aug 2021 to Dec 2022
Connecticut Information Technology Institute (UConn)
Built a Selenium pipeline for 2,000+ client accounts, forecast client-interaction and quiz-completion trends, and deployed a Random Forest binary classifier on Azure ML. Ran A/B tests on landing pages and quiz variants. Delivered Power BI and Streamlit dashboards with SSIS/SSRS reporting, working with the web team to turn findings into design decisions.
May 2022 to Jul 2022
American Century Investments
Built a BERT sentiment-scoring pipeline over GDELT news data for a Tableau dashboard tracking controversy sentiment and article volumes across portfolio holdings. Created an S3 ETL pipeline for MSCI ESG controversy reporting and contributed KPI analysis using REST APIs, SSIS/SSRS, MongoDB, and ANOVA. Compared classical ML baselines with BERT and selected BERT for higher ESG sentiment-classification accuracy.
May 2017 to Jul 2021
Alliant InfoTech Ltd.
Built Azure SQL and PCA-based segmentation with K-Means/DBSCAN, cutting email campaign costs 20%. Python/ETL pipelines halved processing time; A/B tests lifted landing-page conversion 45%, and predictive models improved Q1 retention 30%. Created Power BI dashboards across 15+ KPIs and Excel reporting that saved $78K per quarter. Engineered feature and imputation pipelines and used XGBoost in R to deliver predicted probabilities to the operations team.
Open the two-page resume preview or download the PDF.
If you want to discuss applied ML, data products, or a research workflow that needs more rigor, reach me directly.
kshp95@gmail.com