Main
Peter Shaffery
I am a passionate and curious data scientist interested in novel problems and challenges. I have applied statistical and machine learning models in contexts ranging from higher education to the power grid to ecological systems.
Selected Positions
Network Quantitative Engineer
Meta
Denver, CO
Present - 2022
- Used data analytics and statistical modeling to improve the efficiency of the Meta network backbone
- Apply time series models to forecast demand at multiple time scales
Data Scientist
University of Colorado Boulder
Boulder, CO
2022 - 2020
- Worked directly with CU Boulder admissions and recruitment staff to measure efficacy of marketing interventions, segment and understand prospective student populations, and forecast enrollment and prospect engagement
- Contributed to design of experiments testing interventions to improve enrollment, results enabled re-allocation of a substantial portion of the communications budget
- Built and validated cross-team data resources such as novel datasets and dashboards
- Developed, tested, and documented pipelines using Python and SQL to extract bulk data from admissions CRM for analysis and storage.
Research Intern
National Renewable Energy Laboratory
Golden, CO
2020 - 2019
- Proposed novel Bayesian time series approach to disaggregate solar power generation from gross home power consumption. Proposal improved model error over other state-of-the-art methods by up to 50%
- Contributed code and methods to a project using high resolution, fisheye cameras (“Total Sky Imagers”) to estimate and forecast local solar energy availability
Research Assistant
Dukic Lab
Boulder, CO
2019 - 2015
- Developed and analyzed a novel random matrix model to explain phenomena at the intersection of ecology and epidemiology
- Published and presented at Society of Industrial and Advanced Mathematics conferences (both General and Regional conferences)
- Worked with CU Boulder Office of Data Analytics to forecast graduation rates and tuition revenue, using Bayesian survival models.
Teaching
Instructor, Advanced Statistical Modeling
University of Colorado Boulder
Boulder, CO
2021
Instructor of record for a mixed undergraduate/graduate course covering multiple regression theory, generalized linear models, and elementary Bayesian statistics.
- Adapted existing course examples to demonstrate modern R tools such as
tidyverse
,ggplot2
, andrstanarm
- Created 4 new weeks of course material which included introductory Bayesian statistics, imputation, and causal modeling
Teaching Assistant, Bayesian Statistics and Computing
University of Colorado Boulder
Boulder, CO
2020
Teaching assistant for a mixed undergraduate/graduate course covering Bayesian statistical theory and computation
- Created and presented two lectures on Hamiltonian Markov Chain Monte Carlo and its implementation in
Stan
Teaching Assistant, Calc 1-3, Differential Eq’ns
University of Colorado Boulder
Boulder, CO
2015-2019
Teaching assistant for Calculus 1-3, Differential Equations, and Psychological Statistics
- Graded, lectured, and published course material including worksheets, quizzes, and study guides
- Led “break-out” computer lab sections accompanying Calculus 3, Differential Equations, and Psychological Statistics which introduced students to Mathematica, MATLAB, and R (respectively)
Publications
A note on species richness and the variance of epidemic severity
Journal of Mathematical Biology
2020
Automated Construction of Clear-Sky Dictionary from All Sky Imager Data
Solar Energy
2020
Bayesian Structural Time Series for Behind-the-Meter Photovoltaic Disaggregation
IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT)
2020
Education
PhD., Applied Mathematics
University of Colorado
Boulder, CO
2015-2020
Investigated the statistical dynamics of multi-species epidemics and applied Bayesian time series methods to power grid forecasting problems
B.Sc, Physics and Mathematics
University of Massachusetts
Lowell, MA
2009-2013
Double major in Physics and Math, honors thesis title “Parameter Estimation Consistency Between MCMC and the Fisher Information Matrix”