* Forecasting system for optimizing ads pricing and use of direct response marketing* Budget sizing and optimal pricing tools for job ad sales (fitting a price elasticity model of job clicks and job applies* Data processing and model building in hive and spark

* Brought to production implementations of core ML/data products such as classifying job titles, and contributed our prominent salary prediction features. ▹ End-to-end, from research and data exploration to model building and human testing to optimization for production.* For the search A/B testing platform I 1) advocated for and implemented key statistical components, 2) created a completely new, much-refined UI, and 3) was solely responsible for the system maintenance.* Driven by curiosity and an experimentalists viewpoint, became a substantial expert on all the important practical details of Glassdoor's jobs and employer data.* Contributed many engagement and valuation insights around B2B products, CLV of marketing efforts, and user job search engagement* Became the company Tableau expert and author of a few very high-demand dashboards.

• Developed the web app http://NewsSpectra.com which presents alternative coverage of a topic on a spectrum of readability.• Scraped and extracted information from Google News and individual sources with scrapy, BeautifulSoup, and goose.• Used Python tools nltk and scikit-learn to tokenize, cluster, and rank articles by Flesch-Kincaid reading level.• Published application using AWS, MySQL, gunicorn, supervisor, Bootstrap, and D3.

• Filtered massive datasets (> 1 TB) to find very rare signals using local and world-wide batch computing, contributing to the discovery of the Higgs boson.• Built C++ applications on top of the shared tools of a collaboration of 2000 scientists to calibrate, resolve ambiguities, and filter data.• Developed a framework in Python and ROOT for specifying, building, and publishing plots to the web.• Applied unit tests to publications by factoring out numerical quantities in JSON/YAML+Python.• Controlled and monitored data acquisition, requiring quick reactions and efficient communication with colleagues.

• Used a large stack of Monte Carlo simulations (from particle production to detector response), as well as extrapolations from data, to quantify signal and background.• Increased signal to background separation in selection and identification problems using boosted decision trees.• Computed 95\% confidence intervals for new physical parameters using likelihoods built from data and statistical models of signal and background (with many nuisance parameters quantifying uncertainty).• Managed Monte Carlo simulations by translating colleagues' informal requests into formal job specifications, testing and submitting the jobs, and monitoring the results; built tools in Python to streamline all of these steps.• Contributed 100+ pages to documents describing the experiment's sensitivity to new physics.

• Validated software by broadly and systematically comparing alternative systems for unexpected discrepancies; found and reported on important bugs.• Led a team of five students to complete assembly and testing of large detector components.

• Developed DAQ hardware and software for balloon and satellite experiments in Geophysics (now part of the Dept. of Earth and Space Sciences).• Built simulations of coded aperture imaging with X-rays.