CIA investor backs geospatial data startup
- By George Leopold
- Oct 20, 2016
When the CIA's venture capital arm invests in a technology startup, it tends to draw attention. The latest example is an MIT geospatial data spinout that has demonstrated the capability of generating visualizations on the fly based on billions of data points.
Highlighting the growing importance of big geospatial data for a range of military and intelligence applications, CIA investor In-Q-Tel joined an early funding round for MapD that was announced this week. The startup has developed a database and visualization platform powered by graphics processors, or GPUs, an approach that is gaining momentum as data volumes grow.
In-Q-Tel joined previous MapD investors that include GV (formerly Google Ventures), GPU developer Nvidia, Vanedge Capital and Verizon Wireless.
San Francisco-based MapD (the company name is derived from an MIT research project that morphed into an open source effort called Massively Parallel Database) said it would collaborate with In-Q-Tel to accelerate unspecified features of the startup's GPU-based platform. The database and visualization platform is already touted as leveraging the parallel processing attributes of GPUs to render visualizations up to 1,000 faster than traditional databases.
The ability to accelerate the visualization of big geospatial data is attracting investors such as In-Q-Tel since MapD's GPU-based approach allows users such as intelligence analysts to work with huge data sets at interactive speed. Those capabilities are being used to meld the cyber and physical worlds as geospatial data underpins a new class of geo-location services.
The three-year-old startup "is one of the newly emerging software companies leveraging advanced GPU hardware to deliver very impressive performance improvements in data visualization and data exploration," George Hoyem, an In-Q-Tel managing partner, noted in a statement. "The ability to interact with and visualize billions of data elements in real-time is a transformative capability for our national security partners."
MapD founder Todd Mostak developed the GPU approach as part of his research at MIT's Computer Science and Artificial Intelligence Laboratory. Focusing on database development, Mostak optimized his database to run in the memory of GPUs, discovering that he could create a supercomputing cluster that ran an order of magnitude faster than databases running on traditional processors.
That insight addressed a growing concern that databases are increasingly unable to keep up with huge volumes of geospatial and other imagery data generated by sensors and satellites. In early testing, the Map-D database crunched datasets larger than 1 terabyte, or trillions of bytes.
"The investment by [In-Q-Tel] validates our pioneering approach of using GPUs to power insight discovery across billion-plus row datasets," Mostak said.
The MapD investment is the latest in a string of bets made by the CIA in emerging data analytics startups as it seeks to leverage geospatial data and bridge the gap between real-time and the agency's trove of historical data. Last month, In-Q-Tel announced an investment in Zoomdata Inc., a visual analytics specialist whose platform was recently added to the CIA vendor Amazon Web Services' cloud marketplace.
The Zoomdata investment follows a partnership announced in June with Databricks, which is commercializing a real-time processing engine based on the open-source Apache Spark project.