We design data and machine learning interviews to be more efficient, equitable, and reflective of the actual job.

As a manager:• Developed roadmap to improve how Metromile used driving data to predict risk and build key data science infrastructure; scaled the team from two to seven post-IPO to implement this roadmap• Led a full-stack team of six engineers to make pricing changes affecting $10M in annual revenueAs an individual contributor:• Owned end-to-end data pipeline to calibrate and analyze terabytes of driving data and built models that better predicted risk based on insights from this pipeline• Developed and open-sourced tools to measure racial and gender bias in pricing algorithms, and socialized wider use of metrics for reducing bias (https://github.com/edjzhang/zipbiaschecker)What I wish I had done better: Investing more in monitoring for changes to our input data would have improved how our models held up over time and prevented a costly mistake - lesson learned!Tech stack: Python, Spark, Scala, Snowflake, PostgreSQL, Splunk, AWS, Databricks, Airflow, Docker, Jenkins

I worked at Time by Ping (TBP) while a consultant at SFL Scientific. I list this separately because I was working full-time for TBP during this time and because it inspired my interest in start-ups, experiencing the energy and opportunities for wearing many hats that they bring:• Trained and deployed production NLP models, maintaining unit and load tests, logging and error alert infrastructure, and a pipeline for model re-training• Supported and presented during sales demos and pilots for prospective clientsTech stack: Python, Flask, spaCy, NLTK, MongoDB, Sentry, Loggly, Docker, Aptible

I came to SFL Scientific inexperienced in coding and machine learning, and I'm grateful for opportunities they provided to learn on the job: • Built machine learning models to predict customer retention for a national retailer, future donations for a nonprofit, and patient outcomes from heart sensor data• Created strategic AI roadmaps and proofs-of-concept for Fortune 500 companies in healthcare, transportation, and consumer products• Developed 16 hours of teaching materials for the Chief Data Officer’s use in a machine learning course

I analyzed and communicated data to improve how nonprofits and foundations collaboratively operate:• Streamlined New York City’s emergency food allocation system using geolocated travel data• Prioritized target states for education and advocacy reform based on assessment data• Made recommendations to coordinate funding from seven foundations for national water issues