* Machine Learning Tech Lead of 4-8 Applied Research Scientists working on internal search product.* Delivered $8M in cost savings and productivity improvements.* Worked cross-functionally with product, engineering, and analytics for planning / goaling of ML projects.* Wrote A/B experimentation library improving team velocity 4x.* Joint architect for Search as a Service, setting the team roadmap for the next 18 months.* Utilized a tech stack comprising Python, Pytorch, XGBoost and Hack, built upon aC++ backend, with Presto/Spark used for data pipelines.

* Technical lead in team of 10 research data scientists working in diverse areas of Meta infrastructure.* Capacity forecasting for video conferencing infrastructure.* Modelling video conferencing call quality for identifying regressions and setting SLIs.* Identifying and resolving problematic components in data privacy; saving $2M+ in annual compute and maintenance costs.

* Built credit card fraud classification model used to save $3M annually using GBDT model with H2O.* Prioritized machine learning efforts across 6 different teams.

* Successfully advocated for the adoption of a risky change in CS procedures, resulting in annual savings of $12M with no effect on revenue.* Wrote "False Prophet", an in-house tool for efficient forecasting in TensorFlow.* Wrote internal library to accelerate Spark pipeline development.* Successfully advocated for the adoption of an emergency measure as a permanentsolution, resulting in annual savings of $12 million

Mathematics research and undergraduate/graduate teaching.* Board Member and co-PI on 5.8 million Euro grant. * Published research articles in top international journals.* Invited speaker at over 80 conferences and seminars in over a dozen countries spanning 4 continents.* Organized international conferences.

Conducted research in areas related to homotopy theory. Organized research oriented seminars for the region. Assisted students and postdocs. Spoke on original research in international conferences and seminars.

Improved numerical stability, precision, and speed (∼10x) of one of the most frequently used option pricing algorithms written in C.

* Designed “Double Market”, a 15,000 line Java based client/server application for simulating dishonest auction markets.* Designed an application to monitor and analyze experiments in real time.* Maintained 3 client/server applications totaling over 40,000 lines of code, updated the code and added javadoc documentation.