I lead the teams at Planet responsible for creation of analytics-ready data products. These data products let users focus on their own domain expertise and the insights they can gain, rather than worrying about data cleaning. We produce Analytics-Ready PlanetScope Data, a near-daily, 3-meter, 4-band product which is harmonized to Sentinel 2. It is highly radiometrically consistent and geometrically aligned over time, and includes an advanced cloud mask. We produce Planet Basemaps (https://www.planet.com/products/basemap/), a spatially and temporally consistent product which is optimized for country-level analysis. And we produce Planet Fusion, a version of the Analytics-Ready PlanetScope Data product which includes modeling of unobserved pixels to provide data in every pixel, every day.

Combined engineering manager and tech lead role. I lead a team which builds and delivers Planet Fusion, an advanced analysis ready Earth observation data product. We provide daily, gap-free multispectral images at 3 meter pixel size.

Scaling algorithms to fuse multi-sensor Earth observation data into a daily time series

Deploying and maintaining machine learning models and tools on Planet's planet-scale database of daily imagery

Lead a team which creates maintainable, scalable ETL pipelines on behalf of Civis's clients.

Combined Engineering Manager and Tech Lead role in the Data Science Research & Development department, with responsibilities including design and implementation of machine learning algorithms and infrastructure for the Civis Platform, and creation and maintenance of Civis data science software. In this role, I also guided the growth of other data scientists on my team, did project and product management for my team's work, and coordinated across teams and departments at Civis.Notable accomplishments include• Led a team to create CivisML, the software powering machine learning on the Civis Data Science Platform• Built, expanded, and maintained infrastructure for data science at Civis, including the Civis Python API client and Docker images for data analysis tasks• Led work on software for survey forecasting using Bayesian statistical inference• Performed data analysis for services clients including share-of-wallet modeling for a large B2B supplier and customer value modeling on smartphone app users• Mentored team members, including two reports promoted to Senior Data Scientist and three promoted to Lead Data Scientist• Established and grew educational initiatives throughout the company, including a journal club, programming office hours, and a conference talk-watching group

Write production software tools which make use of machine learning to let data scientists within Civis Analytics and using the Civis Platform work smarter, faster, and better.My major projects in this role include• Additional features for Civis's supervised learning software, including multi-class modeling and out-of-core modeling for large datasets• Improve existing Civis software for heterogenous treatment effects modeling, greatly improving computational efficiency for some model types and reworking the code for better reliability and maintainability

Led development of data analysis software for the South Pole Telescope's polarization-sensitive camera. The SPT is a cosmology experiment designed to make sensitive measurements of the cosmic microwave background. I transitioned analysis software from the IDL language to Python, extended the software to deal with new analysis tasks, and coordinated the primary data analysis for first year data.

Robert R. McCormick Fellow at the Enrico Fermi Institute

Work on hardware, simulation, and data analysis for the ANtarctic Impulsive Transient Antenna (ANITA) experiment.

ANITA is a balloon-borne radio interferometer which recorded many millions of 100-ns radio signals during a flight over Antarctica. I designed, built, and operated an RF testing system for the instrument, developed analysis techniques for understanding the ANITA data, and searched through data to separate a handful of science signals from millions of background events. My analysis was the most sensitive investigation of this dataset, and it proved that our instrument was sensitive to an unanticipated science signal, ultrahigh-energy cosmic rays arriving in the Earth's atmosphere. I wrote tens of thousands of lines of custom C++ code for this project, making heavy use of the ROOT software package.