Data Scientist
2020 - PRESENTGoogle- Created an interpretable model to explain the following actions in the Google News app.
- Improved product dashboards to measure ecosystem diversity.
- Performed analysis of experimental metrics for proposed product changes.
- Developed custom analysis templates to automate analysis of side-by-side comparisons of news ranking changes.
Technologies: R, Python, Data Science, Statistics, Modeling, Experimental DesignData Science Summer Intern
2019 - 2019United Healthcare- Implemented a methodology to estimate the causal effect of treatment for chronic low back pain (lumbago) treatments using propensity score matching and logistic regression to inform about treatment best practices.
- Developed an interpretable recommendation algorithm for selecting the optimal treatment for lumbago tailored to individual characteristics, demonstrating an improvement over standard care in test data and using IPWE for evaluation.
- Cleaned and validated a complicated electronic health record dataset for analyses using SQL.
Technologies: R, Causal Inference, Reinforcement Learning, SQLConsultant
2010 - 2012Mastercard- Designed and implemented the Small Business Insights engine, a data product linking small business data from multiple sources to increase sales of business banking products.
- Developed and analyzed surveys to capture public opinion about prepaid cards.
- Analyzed millions of credit card transactions using SAS and SQL to support consulting projects for retail banking clients.
Technologies: SAS, Netezza, SQL, Statistics
Robert Pehlman
Data Science and Modeling Developer in Chapel Hill, NC, United States
Member since August 5, 2021
Robert is a seasoned data scientist and has worked for Mastercard, United Healthcare, and currently, Google, where he supports the Google News app. He was also a PhD student in statistics at North Carolina State University while researching causal inference, reinforcement learning, and functional data analysis. Robert's passion is solving challenging data problems by building scalable models and engineering new features to improve model performance.
Robert is now available for hire
Portfolio
- GoogleR, Python, Data Science, Statistics, Modeling, Experimental Design
- United HealthcareR, Causal Inference, Reinforcement Learning, SQL
- MastercardSAS, Netezza, SQL, Statistics
Experience
- Modeling 10 years
- Data Science 10 years
- SQL 10 years
- R 10 years
- Statistics 7 years
- Python 5 years
- Reinforcement Learning 4 years
- Causal Inference 4 years
Availability
Part-time
Preferred Environment
R Studio, Jupyter, Windows, Linux, MacOS
The most amazing...
...project I've completed was to estimate the causal effect of different treatments on acute lumbago and to find the optimal treatment sequence for pain relief.
Employment
Experience
- Nonparametric Gaussian Process Estimation R Package with an Application to Functional Linear Modelshttps://github.com/RobertPehlman/Portfolio/
This project aimed to create an R package to do nonparametric Gaussian process estimation. The estimation procedure takes irregular longitudinal data as input sparse, which are assumed to be subject to measurement error, and outputs an estimate for the covariance kernel of the process. An application of the Gaussian process estimation to functional linear models is explored. We examined the computational complexity of the algorithm under differing circumstances.
Skills
Languages
R, SQL, Python, SASTools
R Studio, JupyterParadigms
Data ScienceOther
Data, Modeling, Statistics, Causal Inference, Reinforcement Learning, Experimental Design, Data AnalysisPlatforms
Windows, Linux, MacOSStorage
Netezza
Education
- Progress Towards a PhD in Statistics2014 - 2020North Carolina State University - Raleigh, NC, USA
- Master's Degree in Applied Statistics2010 - 2013Penn State University - University Park, PA, USA
