Data Scientist Specializing in Machine Learning and Predictive Analytics
Data scientist with 6+ years of experience building machine learning systems that turn messy data into measurable business impact. Strong background in statistics, causal inference, and natural language processing, with deep expertise across the full model lifecycle—from problem framing and data pipelines to deployment and monitoring. Passionate about creating interpretable, reliable models and translating data insights into clear recommendations for product and leadership teams.
Concentrated on machine learning, causal inference, and large-scale data systems
Capstone project: built a real-time churn prediction pipeline for a telecommunications partner
Graduate research assistant supporting a study on algorithmic fairness in lending decisions
Built strong foundations in statistical theory and computational methods
Undergraduate research assistant in the Statistical Learning Lab
Lead development of predictive models for customer lifetime value across a portfolio of SaaS clients, improving average revenue forecast accuracy by 18%
Architect feature store and model monitoring pipelines using Python and MLflow, reducing model retraining time from days to hours
Partner with product and engineering teams to design A/B tests and causal inference studies for new feature adoption initiatives
Mentor a team of four data scientists and establish best practices for code review, reproducibility, and documentation
Built risk stratification models using electronic health records and claims data, enabling early intervention for high-risk patients
Designed and implemented an NLP pipeline to extract clinical concepts from unstructured notes, achieving 87% F1 score
Created interactive Tableau dashboards used by clinical leadership to track quality metrics and reduce hospital readmission rates by 12%
Recognized for leading the development of a company-wide model monitoring platform that reduced unplanned model retraining incidents by 40%.
Presented a production-ready churn prediction framework combining gradient-boosted trees with online feature stores
Demonstrated a 23% relative improvement in AUC versus batch baselines across three industrial datasets
Developed a retrieval-augmented system using sentence embeddings and a fine-tuned transformer model to answer clinician questions from EHR notes
Deployed as a REST API with Docker, reducing average answer lookup time from 20 minutes to under 5 seconds
Engineered features from impression logs and built a distributed training pipeline in PySpark
Achieved a 15% lift in offline AUC and enabled real-time bidding through model export to ONNX
Partnered with local nonprofits to analyze service utilization data and guide resource allocation decisions
Built reusable notebook templates to standardize weekly reporting and make findings accessible to nontechnical staff