Highmark Health
Data Scientist Intern
Summer 2026 · Pittsburgh, PA · Healthcare & insurance
Worked as a Data Scientist Intern at Highmark Health, a nonprofit integrated health-services
organization, applying machine learning and analytics to clinical and operational data inside one
of the largest Blue Cross Blue Shield plans in the country.
Model Monitoring
- Implemented a production model-monitoring dashboard for the Predictive Modeling and Data Science Division using R Shiny, BigQuery, and on-premises Apache Spark clusters. The dashboard unified model statistics from two completely different production pipelines without any preexisting dashboard or monitoring setup.
- Computed statistics for each production load date across drift detection (Population Stability Index), feature importance (SHAP explainability deltas and feature descriptions), model-score changes (matched-cohort score deltas and percentile shifts), feature coverage (null-count analysis affecting model scoring), and model descriptions for business users, data scientists, estimated value, and other key context.
LLM Severity Score Model Pipeline
- Built an LLM severity-scoring decision layer for neonatal cases using prior-authorization clinical summaries, assigning integer scores from 1 to 10 across a two-year training period and benchmarking results against an existing high-cost reserving RNN model. In the final 500-case development sample, the system identified more than $10 million in high-cost claims that had previously received low model scores.
- Productionized the scoring pipeline, including NLP-based preprocessing to address personal and irrelevant information, prompt guidelines for clinical factors, inpatient episode trajectory, gestational age, and other signals, internal LLM function calls, and a composite output combining RNN and LLM scores through isotonic regression.
Overshoot Predictive Model
- Crafted a traditional predictive model using LightGBM over two years of big data to classify hospital overshoot days, measuring the gap between the expected discharge date from the authorization admit date and the actual discharge date.
- Used diagnosis likelihood probabilities across different cost buckets alongside other clinical and authorization information to produce both multi-bucket and single-bucket classifications at 80% to 87% accuracy, with the multi-bucket model still performing relatively well and creating cause for further improvement and exploration.
- Google BigQuery
- MLOps & Production ML
- Deterministic LLM Programming
- LLM Sentiment Analysis Scoring
- Concept and Data Drift Detection
ARxChange (TriCap Technology Group)
Data Analyst
Jan 2026 – present · Remote / New York, NY · Financial services · Healthcare data
Data analyst at ARxChange, a New York–based financial-services platform serving the pharmacy
ecosystem. Building reporting and BI tools on top of AWS-hosted client data, surfacing data-quality
insights, and supporting client onboarding across the customer-success organization.
Payment Plan Model Improvements
- Improved the predictive accuracy of the payment-plan modeling strategy by more than 10% using ensemble methods, cross-validation, hyperparameter tuning, and model explainability techniques.
Automated BI Dashboard
- Accelerated operational workflows by eliminating manual review of batch failures. Automated batch-schedule data flows by parsing Redis configurations, writing outputs to Amazon S3 and Redshift, and powering automatically refreshed Amazon QuickSight dashboards.
- Audited client files following backend engineering deployments, identifying and resolving discrepancies to maintain reliability.
- Amazon Web Services (AWS)
- Microsoft SQL Server
- Amazon S3
- Amazon Quick Suite
- Bitbucket
Walmart Automated Consolidation Center (ACC 7377)
Operations Management Analyst Intern
Summer 2025 · Bentonville-area distribution operations
Partnered with Engineering and Operations leadership, Greg Beard (SSE Engineer) and Andri
Sarmiento (Operations Manager), to optimize label placement and printing workflows at ACC 7377,
and to stand up the first live label-quality monitoring system on site.
- Built the site’s first label-quality monitoring system with Power Automate flows and a Power Apps dashboard, scoped to local-site tooling.
- Generated an estimated $6,000 in cost savings in week one (vs. same period 2024).
- Designed the process to be scalable to additional shifts and consolidation centers.
- Validated changes directly with engineering and operations leadership on the floor.
Case study →
- Microsoft Power Apps
- Operations Management
- Project Management
- Looker Studio
- Google BigQuery
Peak Performance Biomechanics
Data Analyst · Biomechanical Data Specialist Intern
Dec 2024 – May 2025 · Pittsburgh, PA · Noraxon IMU / EMG processing
Ran biomechanical and ergonomic analyses using Noraxon's IMU / EMG software, with additional
processing in Python and R, producing client-facing visualizations and reports.
- Processed IMU and EMG datasets in Noraxon; exported signals for further Python / R analysis.
- Built tailored visualizations and reports communicating client progress against KPIs.
- Produced data-driven deliverables that quantified intervention impact and informed next-block programming.
- Analytical Skills
- Noraxon
- Biomechanics
- Problem Solving
- IMU/EMG Data