Experience

 
 
 
 
 
January 2016 – Present
San Francisco, California

CEO

Dgraph Labs, Inc.

Dgraph is the world’s most advanced graph database. It is horizontally scalable, synchronously replicated, transactional and distributed. It can do arbitrarily deep joins, while minimizing network communication and disk seeks, scaling well as cluster size increases. It is a low-latency, high-throughput database, serving complex queries over multiple independent data sources in real-time.
 
 
 
 
 
August 2007 – May 2013
Mountain View, California

Senior Software Engineer

Google

In reverse chronological order:

Led a team of engineers for a project to build low latency, high throughput graph system to consolidate structured data. Replaced various custom backends with this system, allowing better data sharing and lowering engineering resources required to launch new knowledge projects. Google’s Knowledge Graph and OneBoxes (weather, movie showtimes, flights etc.) are utilizing this technology.

Built a knowledge engine which can truly understand user’s query using Freebase Graph, and generate appropriate results. Eventually launched as Knowledge Bar at google.com for listy queries.

Received OC award for successful launch of Caffeine: Google’s incremental indexing system providing 50% fresher results, and the largest collection of web content ever offered by Google.

Built various solutions around tackling size control, index quality and page freshness issues in a system holding >100PB of data.

 
 
 
 
 
June 2006 – January 2007
Mountain View, California

Software Engineering Intern

Google

Ported Bigtable’s RowMutation C++ APIs to Python. Bigtable is a distributed, scalable, proprietary data storage system owned by Google.

Ported Google Localization Framework from relational DB to Bigtable, significantly cutting down latency of data retrieval. Implemented search and logging functionality for various tools in the framework.

Tech Talks

Badger is a fast key-value database, written to provide a Go native alternative to RocksDB. Typical LSM tree based KV DBs are fast in …

A higher level overview of Dgraph.

This talk digs a bit deeper into the design of Dgraph.

Older Projects

Desktop Flickr Organizer

Graphical Flickr Photo Organizer

FlickrFS

Virtual filesystem for Flickr