Portfolio
Selected work
Systems that connect data architecture, production intelligence, and operational decisions.
From foundations to use
These examples are selected for system scope—not just model novelty. Each connects data foundations to production intelligence and a real operational decision.
Healthcare operations
Enterprise data platform to staffing decisions
A forecasting request became a governed data and ML program connecting clinical, workforce, finance, scheduling, and external systems.
Evidence: Probabilistic capacity forecasts supporting staffing decisions across 45 clinics.
Read the case studySemiconductor manufacturing
Production ML for high-value industrial equipment
Containerized data and ML pipelines moved laser telemetry into predictive maintenance, simulation, and operator-facing workflows.
Evidence: Four pending U.S. patent applications across forecasting, digital twins, control, and agentic AI.
Read the case study
Healthcare AI
Prediction before the point of service
End-to-end pipelines combine claims and eligibility data, feature engineering, probabilistic inference, and evaluation for pre-service healthcare workflows.
Evidence: Production-oriented prediction of CPT codes and patient cost sharing.
See current workMethods behind the systems
My architecture work is grounded in statistical practice: define the decision, represent uncertainty honestly, and test how a model will behave outside the notebook.
Full archive
Supporting articles, methods, and technical demonstrations. Formal publications and patents live under Credentials.
Platforms & industrial systems
-
From Insight to Action: Building an Enterprise Data Platform for Urgent Care
How a forecasting initiative grew into an enterprise data platform—and made proactive staffing decisions possible across 45 urgent care clinics.
-
Building A Fab Data Platform
Containerized ML and data pipelines for semiconductor lithography—Fabscape, Docker, and production decision support.
Healthcare & clinical modeling
-
Predicting Urgent Care Demand Under Uncertainty
Why predicting a single patient count is the wrong objective—and how probabilistic forecasts can support better staffing decisions.
-
Medication Regimen Complexity to Predict Patient Outcome
Evaluating medication regimen complexity indices as predictors of clinical outcomes in critically ill patients.
-
Nonproportional Hazards Models for Patient Outcomes
Survival modeling for COVID-19 patients with acute kidney injury—consulting work on CritCOVIDView.
Statistical & scientific work
-
New Methods for the Oldest Time Series of Its Kind
Bayesian state-space models and wavelet analysis for climate signals in a multi-decadal coastal ecosystem series.
-
Mathematical Models of Microbial Growth
Thermal reaction norms for marine microbes under a NASA Space Grant fellowship.
-
An Improved Method for Missing Data in Time Series
Alternative covariance specifications in Bayesian dynamic linear models to improve imputation accuracy.
Experiments & notes
-
Infection News Analyst (LLM RAG Capability Demo)
A personal Streamlit experiment exploring LLM RAG for infection news and CDC monitoring—not a production surveillance product.
-
Making a Github-Hosted Website with Jekyll and Ruby
How this site was built with GitHub Pages and Jekyll.