Productionized Brand Health Tracker data pipeline. The pipeline collects survey respondent data, weights according to demographic groups, and aggregates raw results into brand metrics, such awareness and preference, consumed by marketing teams globally.Introduced pair programming and code review system team-wide to reduce error, standardize code, and facilitate camaraderie & learning.

- Automated reporting to estimate ROI of TV campaigns using statistical model to predict lifetime value; team achieved 187% ROI.- Developed automation to empower email marketers to define audiences through self-serve tooling. Cut support required from data scientists in half. - With marketing teams, brought in over 300k new paying learners. - Defined company-wide key performance indicators for measuring top-of-funnel growth. - Drafted more effective new hire onboarding process. - Built holdout group for measuring email campaign performance.

Marketplace for online courses.- Defined metric for learning engagement, made it a company-level KPI, builtmodel to study drivers, achieved 8% increase through experimentation. Led project to expandtracking beyond video consumption to include quiz, labs, and exercises.- Led effort to establish culture of experimentation & provide tools for easieranalysis & faster iteration.- Built system for identifying and removing bots from experiment results. Upto 60% of traffic had been bots, which was blocking experiment analysis.- Won "wiki hero" award for enabling others through my contributions to the quality and coverage of documentation at Udemy.

- Developed accuracy measurement for Facebook’s content moderation systems, which remove porn, hate speech, etc. from 40M posts per month. - Determined appropriate metrics, sampling distribution, and headcount requirements for measuring accuracy of 15,000 content moderators.- Provided analytics to identify areas for improvement. Team was able to reduce errors by 40% in a year. - Developed model to identify ambiguous cases, which realigned team to focus on improving accuracy in more actionable, straightforward cases.

-­ Natural language processing research at leading audiobook provider.­- Clustered user comments and identified representative samples.­- Built model to extract keywords from Audible’s entire library of books to tagfor search. Experimented with automatic summarization of books.

Projects in deep learning, neural networks, natural language processing, and learning analytics; coursework in machine learning, inference and statistics, and cognitive science.

­- TV & digital advertising network. Used variational auto­encoding to simulate & preserve sensitive data, critical for our forecasting models.-­ Overhauled design of ad forecasting system.

Freelance project, visualizing data flow for an R/GA client.

A project for a nonprofit test-prep academy serving underprivileged youth. Designed and built an individualized student dashboard, including recommended study topics for each student based on analysis of previous assessments. (Python)

• Developed online interactive visual dashboard system used for business intelligence analytics and reporting company-wide. Replacing ad-hoc emailed reports, the dashboard provides automated, up-to-date business metrics such as total subscribers, downloads, and page views.• Created live-updating dashboard of graphics describing how people use nytimes.com, including most popular articles and referrers.• Prototyped visualization-oriented report for ad clients.• Provided ad-hoc and regular reports.

• Member of the Data Science team for a targeted digital advertising firm that specializes in using predictive modeling to identify new customer prospects for major online brands. • Liaised between Data Science and Account Management teams, providing testing, development and data analysis for the former, and reporting for the latter. • Managed an intern, and with his help, used the D3 JavaScript library and R to develop interactive dashboard-style charts to display recent performance of and changes to current online advertising campaigns. • Developed an automated report that uses data from previous campaigns to predict the budget we can take on for clients; it has allowed us to take on much larger budgets than we would have otherwise. • Provided Account Managers with up-to-date information about their accounts and tools for optimizing campaign performance.• Developed quality control and diagnostic tests for our ad targeting systems.• Provided company-wide seminar to educate employees on our ad bidding process.• Enrolled in self-study online course on machine learning.

• Developed a regression model to predict annual driver mileage to sell to insurance companies and other institutional clients.• Developed predictive models to analyze market and identify best client prospects.• Periodically presented strategic insights to company directors to aid in decision-making.• Studied effectiveness of advertising and marketing efforts.• Coordinated with other departments from all over the company, in addition to outside sources, to collect and verify the integrity of data used in analysis.