I am working on the execution platform team, building a high performance database engine in C++.

Building high performance distributed systems in C++. I work on:

• The SQL compiler to generate a distributed execution plan

• The execution framework to run queries on terabytes of data in milliseconds

More specifically, I have:

• Improved scan performance 10x via vectorization to improve cache locality and reduce branch mispredictions

• Built a partitioning scheme for our columnar store to improve on disk locality for range queries

• Improve backend visibility and observability

My main project was to create a geographic index based on the RocksDB storage engine. It accelerated queries involving searching for latitude/longitude points near a point or in a region.

Applying machine learning to networks, specifically a network of 1.1 million financial transactions, to compute embeddings for nodes in the network and predict growth. Embedding dynamic graphs in low dimensional spaces has applications to finance, medicine, social graphs, etc. Under Jure Leskovic’s group, the Stanford Network Analysis Project.

Improved usability of an internal data intelligence tool. This including building a reusable UI element with a nontrivial interface in React. I took on another project to improve backend efficiency because I believed it was higher impact.