 Live from Las Vegas, it's theCUBE, covering AWS re-invent 2018. Brought to you by Amazon Web Services, Intel, and their ecosystem partners. Welcome back to Las Vegas, as we continue our coverage here on theCUBE of AWS re-invent, day two of our three days of coverage. Happy Wednesday to you, wherever you might be watching. We're joined by Yaron Javik, who is the founder and CTO of Iguazio. Yaron, thanks for joining us here on theCUBE once again. Thank you, hi. Yeah, just for folks at home who might be watching or at their office and not familiar with Iguazio, tell us a little bit about the history of the company, what you saw as the need as the founder, and what your primary focus is. So our key focus is delivering advanced services, the same one that you see in the cloud, high performance for real-time analytics. Essentially what we've seen is a gap where you have all the cloud services in the cloud, but when you're fanning into an edge or an on-prem environment, you're usually consuming like IT, you know, VAMs, et cetera. So what we are doing, we're matching the same level of services. We provide serverless functions, AI as a service and manage databases that can run either in the cloud or on-prem or in a federated edge environment to have one consistent application development environment throughout wherever you are. So on the AI side, just you mentioned that, I mean, as you're looking at your client base, your customers and you're introducing this concept now, right? For those who aren't there yet, what do you sell them on, if you will, or what do they want to know? What don't they understand you think generally? Yeah, so, you know, AI and ML, there are a lot of companies solving that problem, okay? Where we master is the notion of real-time AI, okay? What people are looking is into embedding AI into business applications, okay? The traditional notion is you have a data lake, you throw all the data in, then your data scientists go learn stuff, create nice, you know, dashboards in Tableau, great. So what, you know? What people really want is to build recommendation engines, you know, someone is logging into a website, he gets recommendations, so that requires very short latency of response, okay? You're doing front detection in financial applications, so you're feeding a lot of data. You need to make decisions now, okay? You're doing cybersecurity analysis, so you're feeding data from routers and firewalls and switches and you need to act immediately to whatever is happening. You think about retail stores, things like Amazon Go, cameras, examining your behavior, et cetera. I need to respond very, very quickly. Now this is a much harder problem to deliver AI in real-time than it is in sort of a data science workbench or just a batching notion. And traditionally the way people address that problem is by profiling, creating sort of, you know, every time I'm going to see something very similar to that I'm going to go to database, school, compare, and contrast. But the problem is that you need more and more multivariant analysis on objects that keep on updating. You know, my location keeps on changing. If I'm going to stand in front of this store, I need to get this advertisement or if I've just, I've done some purchase with my card and the bank knows my GPS location, it can cross correlate that and know if it's a fraud or not, okay? So there are more inputs going into the decision. This is where we master the ability to ingest lots of data in real-time. Cross correlate that in real-time to generate what's called feature vectors, all those things that make up a decision. Run the decision based on the traditional AI and deep learning algorithms and they act on it. Whether it's respond to a customer request or block some firewall or whatever. And our focus is time to action and the way we're implementing it is using two major components. One is real-time serverless functions which is an open source we're promoting called Nucleo. A second is a real-time database extremely high performance that attaches to those functions and allow and help stitching the data and calculating and getting the results. So that's the general thing we're doing. Yeah, so that idea of the serverless functions with Nucleo, that's really about bringing what you used to in the cloud and bringing that out into the edge which I think we were talking before and that's our thing of focus for a lot of developers I want to use all of the things I'm used to in the cloud where I can just consume them as services and it's quite easy to deal with but then when I come back into the onsite or on onto the edge in this kind of hybrid cloud model I don't actually have access to all those things anymore and I want to. And it's even beyond that because the lambda came from more of like web hooks, use cases, et cetera. Extremely not concurrence, extremely low performance. You're talking about hundreds of milliseconds of latencies. We were talking about like thousand invocations per second. You know, that's sort of the concurrency, single threaded applications. We're talking about real time applications. You know, hundreds of thousands of events per second. We were talking about latency in the range of milliseconds, response times, you have to respond. So we had to build a different serverless. Something that's real time, something that has right, you know, real time access to data, et cetera. So that's originally where Nucleo came in and then you started seeing pool from customers saying yes, but you're also a multi-cloud serverless. And you can also, I can run your serverless on a laptop for debugging. I can run it on a mini edge appliance because this is my enforcement point. I can run it on-prem because I, you know, I'm stuck with some old gear in my on-prem application and this is what made, started making Nucleo very popular in lots of GitHub stores, et cetera. And the fact that we provide it as a fully managed platform, you know, it's open source, consume it, whatever. But when you're using our managed platform, you get security integration with Active Directory, integration with data, logging, monitoring. So it really provides an alternative to Lambda where you need high concurrency and everywhere. You know, edge, cloud, on-prem, but also high performance, high concurrency for those new workloads of real-time analytics. Yeah. So what are some of the things that customers are using the platform to develop on? Like, could you give us an example of someone who's using some of these serverless functions for real-time application? Yeah, so one of the applications is, we do a lot of work with network operators. You know, Verizon is one of our investors, but also working with different other telcos. So we're doing real-time network monitoring across all their firewalls and network equipment, et cetera, to predict the network behavior. So if there's going to be a failure, is it a cybersecurity attack right now? Things like that. The next level that they went into doing is actually remediation. It's essentially rerouting the networks to bypass falls automatically based on the predicted behaviors or stopping some attacks as they occur. So that's one use case. Another use case in financial services and many other places is predictive network operations, is monitoring again behavior of services, et cetera, like in trading platforms. And knowing that there's going to be a latency spike that's going to impact the trading and essentially going and fixing that in order to not lose millions of dollars of trades or real-time tick analytics, you know? Until now, all the financial applications were very sort of event-driven and complex event-driven, not incorporating deep learning, things like that. Now, think that there are many variants, you know, the president is going to tweet something about some company and then it's going to impact the B over or with stock. So the current high-frequency trading algorithms are not designed for that. Now, if you build all those serverless functions that listens on Twitter and news and all those things and they can start cross-correlating that information to a much smarter decision, they fit in the real-time decision of buying and selling stocks into a lot more intelligent decision, you can make more money, okay? Another application, retailers, okay? We're working with locations where they have 1,000 cameras in a single supermarket because they just inspect the shelves to look into inventory levels and eventually they're going to like an Amazon Go model where they actually want to know to track what you're buying, et cetera. So 1,000 cameras in a store, you cannot ship all that bandwidth through the cloud, right? Okay, and this is where it comes to a federated application model where as a developer, the guys that are cloud-born or cloud-first, you know, they know containers and they know APIs and they know that stuff. They don't know how to build a box that sits in a store, okay? This is the other world of VMs and VNICs, you know, they don't care about that. They want APIs, they want Lambda functions, Dynamo, et cetera. So what we're providing is a mechanism where they can develop in the cloud, test, simulate, run CI-CD pipelines, push artifacts to the store to actually go and do the work. And there we have strong partnerships with at least a couple of the major cloud providers. We have co-sailing agreements with Azure. We're working with Google and I assume Amazon will be next, but those two we have strong relations with already. All right, and before we cut you loose, just give me your idea about the show in general here from what you've seen and kind of how you feel about the conversations that you're apartment. Yeah, I was very busy talking to customers all day, so I haven't had a lot of time, I think interesting announcements, you know, they've made the announcement with VMware, I'm still trying to figure out what have they announced? Because, you know, again, we spoke about the fact that the whole idea of cloud is about service obstructions, not virtual machines, not Kubernetes containers, is about using APIs, using serverless functions, using AI work benches that you can develop this new logic. If I'm going to do this VMware on-prem with Amazon, am I going to get all the SageMaker, Lambda, all that on-prem? Or just more of a tactical thing like Azure Stack, like we're bringing you VMs, we're calling it cloud, you know, just for marketing sake? Is that a real cloud service cloud? Okay, I think it aligns with what we're seeing now with the Kubernetes, I think we had some discussion about it, you know, IBM buys Red Hat, you know, Cisco collaborates with Amazon, VMware buys Eptio, Kubernetes is containers, it's infrastructure. We speak to customers, we show them what we do, serverless, you know, AI work benches, database and service. That's the interesting part, that eliminates IT. If you're putting Kubernetes, it perpetuates IT. Now they need to take Kubernetes, tie to their security system, build Spark on top of container, et cetera. Now there's a lot of IT and DevOps work involved, but many customers that need agility, the reason they're going to cloud is not to use VMs, you know? It's to be able to take some lambda functions, some pre-baked services, glue them together and really come fast to market with an application. Yeah, so what we really want to do is just to cloud all the things, I think. Yes. Cloud all the things. Mission accomplished. Yeah, yeah. Well, thanks for being with us. We appreciate the time we're on theCUBE. Good to see you, sir. Thank you. Thank you. All right, back with more here at AWS Reinvent. You're watching it live and we're on theCUBE.