I work on cool ways to aggregate and use user transportation data in Google Maps.

The team I'm on creates the semantic location history data you see in Timeline in Google Maps. My job is to transform & aggregate that data - especially data about public transit, biking, and walking - and create value for Maps users.

When the COVID pandemic hit, I helped build a prototype model to predict crowdedness on transit vehicles. After that model was launched, I took ownership of the project and rebuilt it into a robust, fully machine-learned model. The launch of that model was featured in the Google Keyword blog post below.

I've also worked on a few internal projects related to measuring real-world transportation volumes across the world. Currently I'm working on improving the in-station walking time estimates that Maps shows users in transit directions.

Zenabi was a young data science and software consulting company. I supported three different projects as a data scientist.

One is for a global nonprofit org, where I maintained a model that predicted the likelihood that prospective members will join. I was responsible for this model from end to end - from feature engineering to machine learning to deployment and delivery via API. I also supported other team members on this project in efforts involving email marketing and website content recommendation engines.

The second was for the NYC subway. I prototyped a model that predicts delays on individual lines with minute-level precision. I spearheaded the project and, with two coworkers, pitched our solution to a panel of MTA management and VCs.

The third was for a company that provides loans for small and medium-sized businesses. As the lead and only data scientist on the project, I provided expertise and analysis and began the process of moving their existing approval model from DataRobot to a custom, Zenabi-hosted solution.

I implemented parts of an in-house user event tracking system for Lucidchart in both TypeScript and Scala. I also built ETL pipelines using Python, Scala, and SQL to provide data for and support the efforts of the data scientists.

As a member of the Diagram Automation team, I maintained Lucidchart's CSV import feature. I also supported other diagram auto-creation and auto-layout features, such as AWS Import and ERD Import.

I worked on a scrum team in two-week sprints, participated in frequent code reviews, and wrote unit tests. I wrote mostly TypeScript (including some Angular) and a little Scala.

I used Python to aggregate and clean geographic data about consumers, businesses, and competing branches. I then used regression to isolate key metrics for predicting success of future branch locations.