scale.bythebay.io: Alex Kudriashova Interview
Recording: scale.bythebay.io: Alex Kudriashova Interview
my name is Alex and I work for us to digital so there is the integration process to make sure that the satellite on our company bills feeds their needs and helps to build them analytical application based on the data by us or by some other sources we often use the hash tag called fat data because the data from satellites is really massive data set of about 5 petabytes right now and it's growing by five hundred terabytes every year which means that we need to be very smart about how we approach the plate of procedure how we process the areas which are requested by user because we need to see what we need to provide only the red pixels which are needed to a specifically user can generate so the main main technology issue for our platform is that are they okay so yeah that with many people okay good first who's good okay yeah in our presentation we often say that satellite data is fat data because it's really huge we are talking about a data set of about five to six petabytes which is growing by five hundred terabytes every year which means that we need to be really smart about how we process the data how we store the data and how we assess specifically the areas which a user asked for so we look at this specifically smaller windows of the areas which are very questioned by users yeah so the first day when our website was launched and the browser for assessing de satellite imagery was available we had about several tens of thousands of requests and we figured out where quickly that we need to be smart about what we do and we how are we going to provide specifically smaller areas to the users so this scaling the really issues especially when we are approaching new domains so when we were moving from agriculture to forest management to monitoring of cities to work with finance organizations it meant that this scale of our observed area of interest is really growing and we need to be able to serve all these regions right so we build we deploy our constellation of satellites to image the entire land mass every day in 22 meters in 1 pixel resolution and also we are building a constellation of high resolution satellite for imaging the major locations of the world weekly in 2.5 meters in 1 pixel which means that our person capacity should increase by a factor of 10 [Music] okay so we approached a technology of micro services based on AWS lambdas in order to scale the process so we can do processing up to 10,000 functions in parallel which means for us the processes in that crowbar they're generating a crop map for the entire Corn Belt in about 30 minutes so these ways good for us not only because it is scalable but also it's quick to deploy the in straight startup and need to have really quick iterations to demonstrate the proven concept of what we're doing it's really cool about scale by the bay that people with different backgrounds different stacks of technologists are open to share the approach and another thing which is really cool is that it's not only their high-level talks it's really in depth of different technologies and people are literally like taking videos of every slide and taking shots of previous slide because they're going to use this information about in their work [Music]