● Designed and launched the List API function to fetch fact-based, internet-sourced lists that are seamlessly showcased on Excel or Google Sheets.

● Constructed the account, team overview, settings, and sign-in webpages, using Next.js and Typescript.

● Fine-tuned and validated a BERT base model for our largest client, achieving 80% accuracy on their open ended tagging questions.

● Provisioned training data for a ML model by aggregating 11M+ data points from 11 sources, leading to estimates that our recommendations could reduce incidents, bugs, and postmortems by ~10-20%.

● Conceptualized and built the integration test infrastructure for Google’s ads benchmarking product to increase confidence in test coverage upon code updates.

● Created a tool to catch 10K+ misconfigurations on Google’s cluster management system.

● Designed and developed a Beam pipeline to generate threshold values for alerting on data anomalies for the Datascope data monitoring service.

● Reduced the number of noisy alerts by 90% to increase the significance of each alert that is sent out.

● Increased coverage of data sources by 79% to incorporate support for additional clients.

● Integrated the SortingHat Data Preparation project with Tensorflow Data Validation to improve TFDV’s feature type inference system.

● Revised paper and developed tutorial video for Project SpeakQL to present at SIGMOD 2019.

● Led and conducted several user studies for the SpeakQL interface to gather relevant data regarding the performance of the interface.