Quantitative research scientist and PMP-certified project leader using administrative, longitudinal, survey, and implementation data across public health, nutrition, education, and program evaluation.
I turn complex data into evidence that is transparent enough to audit and practical enough to use. My work emphasizes explicit research questions, defensible estimands, visible data-quality and model diagnostics, reproducible workflows, and clear boundaries between descriptive, predictive, associational, and causal claims.
Public health, nutrition, and biostatistics. A reproducible Python analysis of deidentified NHANES 2017–2018 public-use dietary data. It demonstrates two-day dietary measurement, complex-survey weighting and domain estimation, Taylor-linearized uncertainty, missing-data reporting, descriptive regression, unit tests, deterministic outputs, and scheduled validation.
Methods and limitations · Generated findings
Evaluation analytics and messy multi-source data. A reproducible R and Stata workflow that standardizes, deduplicates, joins, and audits synthetic enrollment, service, outcome, and site data. Reviewers can inspect the quality rules, data dictionary, audit trail, tests, and continuous-integration checks.
Data dictionary · Quality rules
SQL, business intelligence, and decision support. A tested analytics workflow with a SQL metric layer, dimensional model, reproducible Python-generated data, Power BI-ready measures, dashboard previews, implementation-risk monitoring, and an executive decision memo.
Dashboard preview · Metric definitions · Decision memo
| Project | What it demonstrates |
|---|---|
| Public Health Methods Lab | Nutrition epidemiology, respiratory-disease surveillance, direct age standardization, outbreak risk ratios, rolling signals, Kaplan–Meier analysis, automated tests, and generated outputs |
| Nutrition epidemiology case study | Two-day dietary-recall averaging, energy-adjusted fiber and sodium density, completeness reporting, descriptive group contrasts, uncertainty, and measurement-error boundaries |
| Quasi-experimental program evaluation | Propensity-score matching and weighting, common-support and balance diagnostics, clustered inference, regression adjustment, and sensitivity across estimators |
| Multilevel outcomes analysis | Three-level longitudinal modeling, variance decomposition, random effects, contextual variation, interactions, and residual diagnostics |
| Structural equation modeling | Confirmatory factor analysis, measurement invariance, FIML, latent-variable mediation, model diagnostics, and careful noncausal interpretation |
| Project | What reviewers can inspect |
|---|---|
| Evaluation data-quality toolkit | Data contracts, domain/range and cross-field rules, issue-level audit output, reusable SQL checks, and continuous integration |
| Student success predictive modeling | Temporal validation, calibration, capacity-aware thresholds, subgroup diagnostics, model cards, and human-review controls |
| SQL analytics case study | CTEs, window functions, cohorts, anomaly review, tested outputs, metric documentation, and decision-ready interpretation |
| Research project-management toolkit | Project charters, evaluation plans, work plans, risk and stakeholder registers, stage gates, change control, issue templates, and automated template validation |
- Breakfast consumption in low-income Hispanic elementary school-aged children — first-author cross-sectional study of anthropometric, metabolic, and dietary parameters
- Impact of a school-based gardening, cooking, and nutrition intervention on diet intake and quality — TX Sprouts cluster randomized controlled trial
- Design and participant characteristics of TX Sprouts — intervention design and cohort profile
My persistent researcher identifier is ORCID 0000-0002-1140-3185. The complete public publication list is available through My NCBI Bibliography, with an additional profile on ResearchGate.
- Start with the decision and estimand. Define the population, comparison, outcome, time window, and interpretation before fitting a model.
- Make validity visible. Surface missingness, data quality, balance, calibration, clustering, uncertainty, subgroup behavior, and model assumptions.
- Build for reproduction. Use deterministic synthetic data, executable workflows, tests, continuous integration, data dictionaries, and saved reference outputs.
- Communicate limits clearly. Separate descriptive, predictive, associational, and causal claims; keep privacy and responsible-use constraints close to the results.
Methods: dietary recall analysis, complex-survey analysis, surveillance rates, direct standardization, cohort measures, time-to-event analysis, quasi-experimental designs, causal inference, longitudinal and multilevel models, measurement models, missing-data methods, uncertainty, and sensitivity analysis
Analysis: Stata (advanced); R (working proficiency); Python (portfolio workflows); SQL (foundational, including AWS Athena extracts and portfolio projects)
Data and reporting: Power BI (basic); reproducible Quarto reporting; limited Tableau and Snowflake exposure
Delivery: Git, GitHub Actions, automated tests, data contracts, model cards, decision memos, and research governance
Project leadership: PMP-certified project leadership, research operations, stakeholder engagement, scope management, and risk management
Portfolio projects use either deterministic synthetic records or explicitly documented deidentified public-use data. No client, student, patient, protected health information, restricted records, or row-level public-use files are republished. Each project is designed to expose the full workflow—assumptions, code, quality checks, outputs, interpretation, and limitations—rather than only a polished final chart.