Statistics PhD Candidate • Data Scientist • Statistical Machine Learning
Bayesian Modeling • Biostatistics • Survival Analysis • Experimentation • Applied ML • Python & SQL
I build statistically rigorous models and experiments, then turn them into systems that can be evaluated and used.
I am a PhD candidate in Statistics focused on applied data science and statistical machine learning.
My work combines Bayesian modeling, experimentation, uncertainty quantification, and applied ML with the engineering needed to make analyses reproducible and usable.
I work primarily in Python and SQL, and I enjoy turning ambiguous problems into clear statistical questions, evaluating competing approaches carefully, and building practical tools around the results.
Target roles: Data Scientist • Applied Data Scientist • Statistical Data Scientist • Applied Scientist
Programming & Data
Python • SQL • PostgreSQL • pandas • NumPy • SciPy
Statistics & Machine Learning
PyMC • scikit-learn • PyTorch • Bayesian inference • MCMC • Variational inference • Experimentation • Bootstrap methods
Applied ML Systems
FastAPI • MLflow • Streamlit • Git • pytest
Research Areas
Bayesian computation • Spatial statistics • Graph-structured models • Uncertainty quantification
End-to-end statistical modeling system for Arkansas county overdose mortality using CDC WONDER data.
- Developed a suppression-aware Negative Binomial likelihood that incorporates interval-censored death counts instead of dropping or midpoint-imputing them
- Modeled all 75 Arkansas counties across 525 county-year records, including 362 suppressed observations
- Compared pooled and hierarchical Bayesian models using a strict 2018–2023 training / 2024 temporal holdout
- Tracked experiments and model lineage with MLflow, published selected predictions to PostgreSQL, and served results through FastAPI and Streamlit
- Added a grounded AI explanation layer that uses structured API results rather than generating model predictions
Tech: Python · PyMC · PostgreSQL · SQL · MLflow · FastAPI · Streamlit
Experimentation and commercial analytics platform for measuring whether advertising creates incremental business outcomes rather than relying only on attribution.
- Built randomized treatment/control experiments with approximately 80/20 assignment
- Estimated conversion lift, incremental orders, and incremental revenue alongside attributed metrics such as ROAS
- Added analytic confidence intervals for conversion lift and member-level bootstrap uncertainty for order and revenue effects
- Built reusable SQL analytical marts and translated experimental results into campaign-level budget recommendations
Tech: Python · SQL · PostgreSQL · Experimentation · Bootstrap
Research project on flexible spatial dependence for Conditional Autoregressive models.
- Developed graph-spectral extensions of CAR models to relax fixed spatial dependence assumptions
- Implemented collapsed variational inference and MCMC under a common model formulation
- Evaluated recovery, misspecification, and prediction under block-missing spatial observations
- Focused on interpretable Bayesian modeling and uncertainty for structured spatial data
Tech: Python · Bayesian Statistics · CAR Models · Variational Inference · MCMC
A retrieval system for searching internal text and markdown documents by meaning and generating answers grounded in retrieved evidence.
- Built document ingestion, chunking, embedding, and storage using PostgreSQL + pgvector
- Implemented metadata-filtered semantic similarity search
- Built a FastAPI service that returns grounded answers with citations and retrieved chunk snippets
- Designed fallback behavior for weak retrieval rather than forcing unsupported answers
Tech: Python · FastAPI · PostgreSQL · pgvector
- Bayesian and probabilistic modeling for real-world data problems
- Experimentation, incrementality, and uncertainty quantification
- Scalable variational inference for structured statistical models
- Spatial and graph-based statistical modeling
- Building reproducible workflows that connect analysis, evaluation, and usable software
