 Welcome back everybody to AWS re-invent. You're watching theCUBE, the leader in high tech coverage. My name is Dave Vellante with my co-host David Nicholson. We're here all week. We got two sets, 20 plus thousand people here live at AWS re-invent 21. Of course, last year was virtual. We got a hybrid event running. We had two studios running before the show, running a lot of prerecords. Really excited to have Ninshad Bardhuliawala who is the chief product officer at DataRobot. Really interesting AI company. We're going to talk about insights with machine intelligence. Ninshad, it's great to see you again. It's been a while. It is great to see you as well. And I'm so happy to be on theCUBE, I think, eight years since I first came on. Yeah, that's right. When you launched the company that you founded back then, Paxata on theCUBE. Now part of the DataRobot family. Now part of the DataRobot family. And of course, friend of theCUBE, Chris Lynch is the executive chairman of DataRobot. So a lot of connections. I always joke, 100 people in our industry, 99 seats. But tell us about DataRobot. What's the scoop these days? So thanks very much for the opportunity to speak with both of you. I think we're seeing some very interesting trends. We've all been in the industry long enough to recognize that hype cycles, they're cycles, they go in waves. And the level of interest in AI has never been higher. Every company in the world is looking for the opportunity to take advantage of AI to improve their business processes. Whether it's to improve their revenue, it's to lower their cost profile, or it's to lower their risk. What we're seeing that's most interesting is that we spend a lot of time working with companies on what we consider applied AI. That is, how do we solve real business problems with the technology? And not just run a bunch of experiments. It's very tempting for a lot of us, Dave and David, to spin up a spark cluster with 10,000 nodes and slosh a bunch of data through it. But the question we always ask at DataRobot is, what is the business value of doing this? Why are we using these AI techniques and in order to solve what problem? So the biggest trend we see at DataRobot and one that we feel we're very well positioned to solve is that companies are coming out of that experimental phase. There's still a lot of experimentation going on and they're saying, okay, we stood up a cluster, we got a bunch of Python notebooks running around here, but we haven't really seen a return on our investment yet. DataRobot, can you help us actually make AI real and concrete in terms of achieving a specific business outcome for us? Well, and I want to test something on you, Nenshott. Something we've talked about a lot on theCUBE is a change in the way in which companies are architecting their data. When we first met, it was like, okay, create a Hadoop cluster and then Spark came along to make that easier, but it was still this highly technical, highly centralized, hyper-specialized roles where the business people who have a really good understanding of the outcome had to kind of beg to get what they wanted because it was so technical and success was defined as, hey, it worked, we ran the experiment and it looks like it has promise. So now it seems like companies like DataRobot, you're democratizing AI, allowing organizations to inject AI into their business processes, their applications, and it seems to be more business-led. I wonder if you could comment on that. I think that is a very astute observation. We launched this concept a little bit earlier this year of AI Cloud, and the idea behind AI Cloud is if you want to democratize AI, which is in fact has been DataRobot's vision since 2012, we were the first company on the cloud, the first AI Cloud that ever existed was DataRobots in 2014, and the entire idea was that we knew that data scientists would always play a very important role in an organization, but yet the demand for AI would vastly outstrip the supply, and so in order to solve that challenge, we built AI Cloud, we've actually spent over a million engineering hours in building this technology over the last decade, and put this together in a way where all of the different personas in the organizations, you have people who create AI applications, those are the folks we usually think about, but those are the data scientists, those are the analysts, those are the data engineers, but then you actually have to put it into production, you've got to run this system, so you also have to democratize this capability for the folks who are going to operate the system, for the folks in risk and compliance who are actually going to ensure that the system is operating in accordance with your policies and compliance regimes, and then the third wave of democratization, which we've just embarked on, is then how do you bring AI into the hands of the actual business people? How do you put on a mobile device or a web browser or in context in an application with a decision the ability for AI to drive a decision in your organization which leads to an action which helps drive you towards the outcome you're trying to optimize for? So AI cloud is about this pervasive tapestry bringing together the creators, the consumers, the individuals who operate these systems into a single system that can lower the barrier to entry for people who don't have the skills, but allow you to plug in and go deep underneath the covers and modify whatever you need to if you have that level of technical skill, and that ability for us to kind of slide the slider in one direction or the other, I can slide it to the right and say, I want all automation, something DataRobot has pioneered and is absolutely the leader in. But we can also, especially in these last couple of years, say I want to be able to use as much code as I want to bring in, and the beauty of the model is that customers can choose how much they want to let the machine drive or how much they want to let the human being drive. David, I love that idea of a slider because now you're talking about generalists getting access to really powerful tooling. Yeah, now exactly, and I'm curious, what's your view on where we are culturally with AI at this point? What I mean by culturally is the idea that, okay, that's great, you put powerful tools in the hands of business users. Do most of us still need to have a lot of visibility under the covers to understand the inner working so that we trust what we're being told? I'm fine pulling a lever and having a little biscuit come out of slot, as long as I've gotten a tour of the kitchen at some point in time. I mean, where are we with that? Where's the level of trust? Absolutely, fantastic question, and it's one that's actually pervasive to the way DataRobot operates. So trust gets engendered by multiple different capabilities that you build throughout the platform. The first one is around explainability. So when you get a prediction from a system, just like you mentioned, if the stakes are not very high, you know, we're here in Las Vegas, so of course I'm thinking of slot machines. If you get a biscuit at the end of it and it tastes pretty good, hey, great, right? When you're making a mission-critical business decision, you don't want to be in the position where you don't understand why the system is making the decision it does. So we have historically invested an enormous amount of effort in explainability tools. Having the system actually at a prediction level explain to you why is it making the recommendation it's making. For example, the system says this customer has a high likelihood of churn. Why? Because their account balance has been declining over the last five months. Number two, because their credit score has been going down, and what gives you the trust is actually the machine and the human able to communicate in the same language and same vernacular about the business value. So that's one part of it. The second part is about transparency, right? So one of the things that the automated machine learning movement that DataRobot pioneered has been, I'd say rightfully criticized for, frankly, is that it's too much of a black box. It's too much magic. I load my data set, I press the start button, and DataRobot does everything else for me. Well, that's not very satisfying when you have a 10 or $100 million decision coming on the other side. Even if the technology is actually doing the job correctly, which DataRobot usually does. So where we've morphed and evolved our position in the market, and where I have driven our technology portfolio at DataRobot is to say, you know what? There is a very important aspect of trust that needs to be brought to bear here, which is that if somebody wants to see code, let them see code. And in fact, the beauty of AI Cloud is that on the same platform, the people who don't like code, but are very good at understanding the business domain knowledge in the context, they now have the ability to do that. But when they're at the stage, before they're going to deploy anything to production, now you can raise your hand at DataRobot and actually use our workflow and say, I need a coder to review this. I want the professional data scientist who has all this knowledge, who understands and has read up on the latest advances in hyperparameter tuning to look at the model and tell me that this is going to be okay. And so we allow both the less technical folks and the very deep technical data scientists the ability to collaborate on the same environment, which allows you to build trust in terms of the human side of, hey, I don't want to just let anybody throw a model into production. Yeah, I like, I mean, I see those, the transparency and the explainability is almost two sides of the same coin. That's right. Because, you know, if you're going to be accused of gender bias, you can say, no, here's how the system, it's not like, you know, you think about the internet, it tells you it's a cat, but you don't really know how the machine determined that. Right. You're breaking apart, blowing away that black box. And the other thing I like what you said was, you have data producers and data consumers. And you also talked about context, because a lot of times the data producers, they don't necessarily care about the context or the data pipeline people. That's right. They don't necessarily care about the context. So, okay, so now we're at the point where you're democratizing data, you're doing some great work. What are some of the blockers that you see today that you're obliterating with data robot? Maybe you could talk about that a little bit. Sure. So I think, you know, one very important concept is that in a democracy, we talked about democratization, you still have rules. You still have governance. It's not a free for all. But the free for all version of that is called anarchy. That's not what any company wants. So we have to blend the freedom and flexibility that we want businesses to have with the compliance and regulatory observability that we need in order to be successful. So what we're seeing in our customer base and what companies are coming to Data Robot to discuss is, okay, we've tried these experiments, now we want to actually get to real business value. And one of the things that's really unique about Data Robot is that we have worked in our system on over one million projects training models inside Data Robot. We have seen every type of use case across different industries, whether it's healthcare or manufacturing or retail, we have the ability to understand those different data sets and actually to come up with models. So we have that breadth of information there. If you aggregate that over time, right? So again, we did not come to AI. This is not a fad for us. We didn't start as one kind of company, then slap the AI label on and say, hey, we're an AI company now, right? We have been AI native since day one. And in that process, what we have found is working on these million plus projects on these data sets across these industries, we have a very good sense of which projects will actually deliver value and which don't. And that gets to a previous point that you were making, which is that you have to know and partner with an organization who it's not just about the technology. So we have fantastic people who we call our customer facing data scientists who will tell the customer, look, I know you think this is a really high value use case, but we've tried it at other customers and unfortunately it didn't work very well. Let's steer you, because you need with a technology that is largely at the early stage and the maturity that organizations have with it, you need to help them in order to deliver success. And no vendor has delivered more successful production deployments of AI than data robots. Yeah, so you can tell them, don't count down that path, it's a dead end, it's a cul-de-sac, so just avoid it. So we talked about transparency, explainability, governance. Can you get that to the point where it's self-serve? As you put data in the hands of business people where the context lives, the domain experts, can you get to self-serve and federate that governance? Yes, you can. That's one of the key principles of what we do at Data Robot. And it comes back to a concept that I learned, you both will remember, we were in the Sarbanes-Oxley crazy world of, I don't know, was that 15 years ago? Save data warehousing. Everybody wanted to talk about socks. My wife would hear me on the phone, she'd be like, what is your sudden obsession with socks? I'm like, no, no, it's not what you think. But what came from Sarbanes-Oxley are these long-standing principles around the segregation of duties and segregation of responsibilities. You can have democratization with governance if you have the right segregation of duties. So for example, I have somebody who can generate lots of different models, right? But I don't allow them to, in a self-service way, just deploy into production. I actually have a workflow system which will go through multiple rigorous approvals and say, these three people have signed off, they've done an audit assessment of this model, it's good to go, let's go and drop it into production. So the way that you get to self-service with governance is to have the right controls and policies and frameworks that surround the self-service model with the right checks and balances that implement the segregation of duties I'm talking about. And you get that right and then you can automate it and then you can really scale. That's right. We've got to have you back because it's such a great topic. We barely scratched the surface. It was great to see you again. Congratulations on all the success. And as I say, anytime, let's do this again. Fantastic, thank you so much. All right, you're welcome. And thank you for watching, we're watching theCUBE's coverage of AWS re-invent 2021. Dave Vellante for David Nicholson. Keep it right there. You're watching theCUBE, the leader in high-tech coverage.