Prakshep – Using pictures of our world to weave stories

The mandate of every business is to always stay informed about itself. In an increasingly complex and interconnected world, making business reality available at zero degrees from the decision-making process remains the demand for most enterprises today. In the absence of a single common backbone of reference, or any historical perspective or dynamicity in their information, organizations today are forced to build disparate “versions-of-truth” depending upon human inputs, empirical knowledge, crowd-corrected assumptions etc.

Prakshep brings Businesses their own reality by providing a comprehensive view about all the things on earth we are interested in: people, buildings, crops, roadways, water, ships, coasts, cities and villages. Utilizing Computer Vision Intelligence deployed parallelly on top of one of the largest repositories of Med-Res Sat-Imagery to rapidly detect and classify the evolving ground realities, and by leveraging all available and disparate context data from the larger world, such as market reports, ground-captured images, price movements, geo-political conditions, location streams, social sentiments etc., Vama, resident AI, is able to quickly learn about things humans are interested in.

Vama utilizes a lean structure called Thorium for the delivery of these real-time insights from the massive underlying dataset. At the bottom of the pyramid lies about 300 Peta-bytes of Medium Resolution Satellite-imagery exponentially expanding: the Landsat-8 satellite images the entire surface of the earth every 16 days. This Massive Data is held offline, or on cloud-distributed glaciers, while around 10 Tb of Master Intelligence is extracted to be kept at a nearer business recall on-disk, constituting a Base Map of our world, and from this is extracted the final essence that resides in SAP HANA: real-time insights that can be delivered on-the-fly to businesses: via humans and machines alike, available on-device, on-premise or on-cloud.

Prakshep leverages SAP HANA to make Massive geo-spatial data available for deeper machine understanding in a scalable and economic manner, and the resulting insights – fast, frequent and reliable, can be directly integrated into any existing business systems and workflows via APIs. SAP HANA enables multi-line and cross-embedding of multiple intelligent models, and provides a computational leverage that represents a technological singularity in terms of crunching massive geo-spatial data: a single run of pan-India analysis takes Vama 19 minutes to deliver, while earlier non-hybrids used to take days, if not months, for the same. This kind of disruptive time-economics also brings complex Sat-imagery analysis and insight delivery into the realm of continuous human interaction: HANA speaks SQL, and talking to our ultra-fast HANA Db is simple for anybody who talks SQL, man or machine.

Prakshep is democratizing a common Sat-backbone of Insights by making basic earth information, in the form of Continuous Digital Surveys, available for anyone to use every 15 days, via channels, partners and reports, both online and offline. Case in point is our India Wheat Production Report for the Rabi Winter 2016-17 season, part of our India Food Story, where we analyzed Med-Res Satellite Imagery for India to predict her total Wheat Cultivation Area, Potential Output & Yield after Losses, almost 100+ days in advance of any available market estimate. These numbers are now validated accurately within 1% and as the First Predictor of Long Cyclic High for Wheat in India.