Adriana E. Reyes
Modeling Healthcare Outcomes Using Real-world Data
New to Contra
Adriana E. is ready for their next project!
Property Valuation ML Pipeline Development
0
0
AI coding agents are fast. They're also great at accumulating redundant functions, ignoring pipeline integrity, leaking features in ways that look fine until they don't, and writing code that technically runs but nobody could maintain.
This is a human review, by a practitioner with 6 years of healthcare data science experience, specifically for codebases where agents have been doing the building.
https://contra.com/products/UXcU8hy7-human-code-review-data-science-and-machine-learnig
2
65
vbcbench.com (http://vbcbench.com) is a benchmarking tool I designed and built end-to-end that scores the transparency of public outcome claims made by value-based care organizations. Think of it as a Moody's-style rating for healthcare outcome reporting.
On the back end, I built an AI-assisted extraction pipeline that pulls outcome claims from public sources (company websites, CMS data, PubMed, SEC filings), applies semantic deduplication, and scores each claim against a deterministic rubric evaluating source accountability, comparator presence, denominator clarity, and timeframe documentation. The scoring engine is fully rules-based (v1.0, scale 1–5) to ensure reproducibility and auditability.
On the front end, I handled deployment on Vercel including SSL configuration and subdomain routing, and iterated on the UI to surface scored results in a clean, interpretable format for a non-technical audience.
The current dataset covers 10 companies and ~51 validated claim snippets, with infrastructure in place to scale across the broader VBC market. Still a work in progress.
1
59
First Author & Data Scientist | Presented at the American Epilepsy Society Annual Meeting
Led the quantitative analysis of epilepsy's multidimensional impact on Veterans' lives using the Personal Impact of Epilepsy Scale (PIES). Built regression models to identify patient and clinical characteristics most strongly associated with seizure, medication, and comorbidity burden and quality of life outcomes. Developed all data visualizations in R and ggplot2. Key findings included a consistent association between seizure frequency and negative epilepsy impact, and age effects aligned with broader quality-of-life literature.
1
40
Built a modular, scalable ML solution for property valuation in Chile as part of a one-week Bain consulting challenge. Starting from a raw Jupyter notebook, I refactored the code into a professional pipeline MVP with a training workflow, FastAPI prediction service, and Dockerized deployment. The repo was designed for future database integration and scalable business needs, showing my ability to transform exploratory work into a production-ready solution quickly and clearly.
1
47
I was the lead data analyst on a peer-reviewed cardiology study with Dr. Kevin Shah, examining disparities in COVID-19 clinical trial enrollment across sex, race, and ethnicity. Using R, I performed descriptive analysis and built multivariable models on 14k+ hospitalized patients from the AHA COVID-19 CVD Registry. The study revealed lower enrollment for Black and female patients, highlighting equity gaps in clinical research and the value of rigorous healthcare data analysis.
0
49