▪ Work on GTSfM at Prof. Frank Dellaert's lab▪ Develop a distributed system in Dask to parallelize GTSfM computations across multiple machines

▪ Team: Google Cloud Networking, Engineering Productivity▪ Built a flag synchronization library, preventing flag mismatch related production incidents▪ Wrote a pull request synchronization feature that aligns the release's PR number with the load tests▪ Refactored common production flags into a centralized location and utilized those in an emulated service▪ Implemented a library to split flags based on flavors and regions, enabling load testing with selected flags▪ Improved usability and maintainability of the load testing framework▪ Technology stack: Python, GCL, Borg, Borgspec, Prodspec, Stubby API, RPC

▪ Built end-to-end personalized generators for the Facebook Group Reels distribution▪ Developed a user interest based Group Video generator for the Facebook Watch Tab▪ Implemented originality controlled Group Video generator for In-Feed Recommendations▪ Technology stack: Python, Hack (PHP), SQL, GraphQL

▪ Helped develop AI automated financial advising application▪ Researched machine learning models for long term market (stocks, bonds, funds) simulation▪ Created Jupyter notebooks to investigate and compare different portfolio management techniques▪ Developed APIs to integrate research results in production using Flask▪ Technology stack: Python, Scikit-learn, SciPy, PostgreSQL