 Okay, welcome back everyone, it's Cube's coverage here in Las Vegas for Amazon Remars. Machine learning, automation, robotics, and space. I'm John Furrier, host of theCUBE, got a great set of guests here talking about AI. Jason Montgomery, CTO and co-founder, Manitium and Ryan Sevig, CEO, founder. Guys, thanks for coming on. We just chatting, lost my train of thought because we were chatting about something else in your history with DataRobot and your backgrounds as entrepreneurs. Welcome to theCUBE. Thanks for having us. Thanks for having us. So first, before we get into the conversation, tell me about the company. You guys have history together, multiple startups, multiple exits. What are you guys working on? Obviously AI is hot here, it's part of the show. M is Mars, machine learning, which we all know is the basis for AI. What's the story? Yeah, really, we're here for two of the letters in Mars. We're here for the machine learning and the automation part. So at the high level, Manitium is a no-code AI application development platform. And basically anybody could log in and start making AI applications. It could be anything from just texting it with the Twilio integration to tell you that you're doing great or that you need to exercise more to integrating with Zendesk to get support tickets classified. So Jason, we were talking too about before we came on camera about the cloud and how you can spin up resources. The data world is coming together and I like to see two flash points. The I call the 2010 big data era that began and then failed, Hadoop crashed and burned. Then out of the woodwork came data robots and the data stacks and the snowflakes. And now you have that world coming back at scale. So we're now seeing a huge era of I need to stand up infrastructure and platform to do all this heavy lifting I don't have time to do. That sounds like what you guys are doing. Is that kind of the case? That's absolutely correct. Yeah, and typically you would have to hire a whole team. It would take you months to sort of get the infrastructure automation in place. The DevSecOps, DevOps pipelines together and to do the automation to spin up, spin down, scale up, scale down. Requires a lot of special expertise with Kubernetes and a lot of the other data pipelines and a lot of the AWS technologies. So we automate a lot of that. So if DevOps did what they did, infrastructure as code, data as code, this is kind of like that. It's not data ops per se, is there a category? How do you see this? You can say data ops, but that's also, it's DevOps, DevSecOps, there's a lot going on. It's not just saying AI ops. There's a lot more. What would you call this? It's a good question, I don't know if we've quite come up with a name. It's not data ops, it's not like- We call it AI process automation. There you go. It's APA, it's not what RPA promised to be. Yes, exactly. But what's the challenge? The number one problem is, I would say, I'm not so much all on heavy lifting. It's a lot of heavy lifting, that for sure. What's involved? What's the consequences of not going this way? If I want to do it myself, can you take me through the pros and cons of what the scale, the scope, the scale of without you guys? Yeah, historically you needed to curate all your data, bring it together and have some sort of data lake or something like that. And then you had to do really a lot of feature engineering and a lot of other sort of data science on the back end and automate the whole thing and deploy it and get it out there. It's a pretty rigorous and challenging problem that there's a lot of automation platforms for, but they typically focus on data scientists. With these large language models we're using, they're pre-trained, so you've sort of taken out that whole first step of all that data collection to start out and you can basically start prototyping almost instantly because they've already got like six billion parameters, 10 billion parameters in them, they understand the human language really well and a lot of other problems. I don't know if you have anything you want to add to that, Ryan, but yeah. I think the other part is, we deal with a lot of organizations that don't have big IT teams and it would be impossible, quite frankly, for them to ever do something like deploy Textract as an example. They're just not going to do it, but now they can come to us. They know the problem they want solved. They know that they have all these invoices as an example and they want to run it through a Textract and now with us they can just drag and drop and say, yeah, we want Textract and we want it to go through this and that's what we want. And expertise is a huge problem and the fact that IT's changing too, right? Put that out there. You guys say, you know, cybersecurity challenges because we have a background on that, so you know, all the cutting edge. So this just seems to be this IT, I hate to say transformation because that's not the way I'm looking for. I'd say stuck in the mud kind of scenario where they have to get bigger faster and the scale is bigger and they don't have the people to do it. So you've seen the rise of managed service. You mentioned Kubernetes. I know this young 21 year old kid who's got a great business. He runs a managed service for Kubernetes. Why? Because no one's there to stand up to clusters. Yeah, it's a big gap. So you have these sets of services coming in. Where are you guys fit into that conversation if I'm the customer? My problem is what? What is my problem that I need you guys for? What does it look like? Describe my problem. Typically you actually, you kind of know that your employees are spending a lot of time, a lot of hours. So I'll just give you a real example. We have a customer that they were spending 60 hours a week just reviewing these accounts payable invoices. 60 hours a week on that. And they knew there had to be a better way. So. The annual review? Like when we got their data, they were showing us these invoices and they had to have their people circle the total on the invoice, highlight the customer name. That person quit the next day. No, like they, hey, you know, they have four people doing this. They're four people doing this. And the point is, is they come to us and we say, well, you know AI can just, basically using something like TextRack can just do this. And then we can enrich those outputs from TextRack with the AI. So that's where the transformers come in. And when we showed them that and got them up and running in about 30 minutes, they were mind blown. And now this is a company that doesn't have a big IT department. So. But they had the ability to quantify the problem. They knew, and in this case, it was actually a business user. It was not a technical. Yeah, the tower. Consequences that's wasted manual labor wasted. And to their point, it was, look, we have way more high valuable tasks that our people could be doing than doing this AP thing that takes 60 hours. And I think that's really important to remember about AI. I don't think it's going to automate away people's jobs. What it's going to do is it's going to free us up to focus on what really matters and focus on the high value stuff. And that's what people should be doing. I know it's a cliche. I'm going to say it again, because I keep saying it, because I keep saying it for people to listen. The bank teller argument always was the big thing. Oh, yeah, they're going to get killed by the ATM machine. No, they're opening up more branches. That's right. So it's like, come on, people, let's get over that. So I definitely agree with that. Then the question, next question is, what's your secret sauce on the customer? I'm going to like that value proposition. You make something go away. It's a pain relief. Then there's the growth side. Okay, you can solve some problems. Now I want the vitamin. Got the aspirin and I want the vitamin. What's the growth angle for you guys with your customers? What's the big learnings once they get the beach head with problem solving? I think the big one is to say that we start with the account payable thing. Because it's so, our platform is so approachable, they go in and then they start tinkering with the initial, we'll call it a template. So they might say, hey, you know what? Actually in this edge case, I'm going to play with this. And not only do I want it to go to our accounting system, but if it's this edge case, I want it to email me. So they'll just drag and drop an email block into our canvas and now they're making it their own. So this is the no code, low code situation. They're essentially building a notification engine under the covers, they have no idea what they're doing. But they get the value. They just know that, hey, you know what? When the amount's over $10,000, I want an email. They know that's what they want. They don't know that's a notification engine, right? Of course, that's called value emails. Yeah, that's exactly what I wanted. All right, so tell me about the secret sauce. What's under the covers? What's the big, big scale valuable secret sauce? I would say part of it, and honestly, the reason that we're able to do this now is transformer architecture. When the transformer papers came out, and then of course, the attention is all you need paper. Those kind of unlocked it and made this all possible. Beyond that, I think the other secret sauce is we've been doing this a long time, so we know where the pain points are. We live through those pain points because we weren't data scientists or ML people. Yeah, you walked a snow with no shoes on in the winter. That's right. He's now got boots on. They're all happy. They've installed machines. You've loaded OSs on top of our rock switches. Yeah, I mean, it's unbelievable how awesome it is right now to be a developer. And now a business user is doing the low code. If you have the system architecture set up, so back to the data engineering side, you guys had the experience got you here. This is a big discussion right now we're having on theCUBE and many conversations. Like the server market, you had that go away through Amazon. Google was one of the first, obviously, to board. But the idea that servers could be everywhere. So the SRE role came out, the Site Reliability Engineer, which was one guy, or gal, and zillions of servers. Now you're seeing the same kind of role with data engineering. And then there's not a lot of people that fit the requirement of being a data engineer. It's like, it's very unique because you're dealing with a system architecture, not data science. So start to see the role of this new persona because they're taking on all the manual challenges of doing that. You guys kind of replaced that, I think. Well, do you agree with it about the data engineer, first of all? I think, yeah, well, and it's different because there's the older data engineer and then there's sort of the newer cloud aware one who knows how to use all the cloud technologies. And so when you're trying, we've tried to hire some of those and it's like, okay, you're really familiar with old database technology, but can you orchestrate that in a serverless environment with a lot of AWS technology, for instance? And that's hard. There's not a lot of people who know that space. And there's no real curriculum out there that's going to teach you how to handle ETLing and also like I got, I want to stream data from this source, I'm using SQL, I'm going to put it all together. Yeah, so it's a lot of just trying to figure that out. So with the large language models too, we don't have to worry about some of the data there too. It's already codified in the model. And then as we collect data, as people use our platform, they can then curate data they want to annotate or enrich the model with so that it works better as it goes. And then kind of curating, collecting the data as it's used, so as it evolves, it just gets better. Well, you guys obviously have a lot of experience together. Congratulations on the venture. Thank you. What's going on here at ReMars? Why are you here? What's the pitch? What's the story? Where's your, you got two letters, you got the M for the machine learning and AI and you got the A for automation. What's the ecosystem here for you? What are you doing? Well, I mean, I think you kind of said it right. We're here because the machine learning and the automation part, but more widely than that. I mean, we work very, very closely with Amazon on a number of fronts, things like text track, transcribe, Alexa, basically all these AWS services are just integrations within our system. So you might want to hook up your AI to an Alexa so that you could say, hey, Alexa, tell me updates about my LinkedIn feed. I don't know, whatever your heart's content is. Well, what about this cube transcription? It's going to be live. Yeah, exactly, 100%. Yeah, we could do that. You don't feed all this in there and then we could do summarization of everything that was on here. Q&A extraction, yeah. And say, hey, these guys are technical. Yeah, there you go. Oh, they mentioned Kubernetes. We didn't say serverless. Chef, puppet, those are words, right? You know, linguistics matters, right? You're getting into, that's a service that no one's ever going to build. Well, and actually, on that point, really interesting, we work with some healthcare companies and when you're basically, when people call in and they call into the insurance, they have a question about their, like, is this going to be covered? And what they want to key in on are things like, I just went to my doctor and got a cancer diagnosis. So the relevant thing here is they just got this diagnosis. And why is that important? Well, because if you just got a diagnosis, they want to start a certain triage to make you successful with your treatments because obviously there's an incentive to do that. Yeah, time series matters in data. Exactly. And machine learning reacts to it. But also it could be fed back old data. It used to be time series, just store it. But now you could reuse it to see how to make the machine learning better. You guys doing anything around that, how to make the machine learning smarter, doing look backs or maybe not the right word, but because you have data, how about you look back at it? That's how it happened. So part of our platform and part of what we do is as people use these applications, to your point there's lots of data that's getting generated. But we capture all that and that becomes now a labeled data set within our platform and you can take that labeled data set and do something called fine tuning, which just makes the underlying model more and more yours. It's proprietary the more you do it and it's more accurate usually the more you do it. So yeah, we keep all that. I want to ask your reaction on this, it's a good point. The competitive advantage and the extra property is going to be the workflows. And so the data is the IP. If this refinement happens, that becomes an intellectual property. That's kind of not software. It's the data modeling, it's the data itself is worth something. You guys seeing that? Yeah, and actually how we position the company is man team is a control plane and you retain ownership of the data plane. So it is your intellectual property. It's in your system, it's in your ADS environment. That's not what everyone else is doing. Everyone else wants to be the control plane and the data plane. We don't want to own your data. We don't. It's a compliance and security nightmare. Yeah. Let's be real. It's like a question, what do you optimize for? Right. And I think that's a fair bet given the fact that clients want to be more agile with their data anyway. And the more restrictions you put on them, why would that, this only gets you in trouble. I can see that being, and plus lock-ins going to be a huge factor. I think this is coming fast. I know what it's talking about in the press, but everyone's like run to silos, be a silo. And that's not how data works. No. So the question is, how do you create siloing of data for say, domain specific applications while maintaining a horizontally scalable data plane or control plane? That seems to be kind of disconnected. Everyone wants to lock in their data. What do you guys think about that, this industry transition we're in now? Because it seems people are reverting back to fourth grade. And to, you know, back to silos. Yeah, I think, well, I think the companies probably want their silo of data, their IP. And so as they refine their models and we give them the ability to deploy it in their own SageMaker, in their own VPC, they retain and own it. They can actually get rid of us and they still have that model. Now they may have to build, you know, a lot of pipelines and other technology to support it, but- Well, your lock-in is usability. Exactly. And value. Yeah. Value proposition is the lock-in. Bingo. Exactly, yeah. You say, hey, more value, I want to get rid of it. Valuable, I'll pay for it. Right. As long as you have multiple values step up and that's what cloud does. I mean, I think that's the thing about cloud that's going to make all this work, in my opinion. The value enablement is much higher. Mm-hmm. So good business model. Anything else here at the show that you observed, that you liked, that you think people would be interested in? What's the most important story coming out of the holistic of you zoom up and look at remars? What's coming out of the vibe? You know, one thing that I think about a lot is we have Artemis here, humanity hopefully is soon going to be going to Mars. And I think that's really, really exciting. And I also think when we go to Mars, we're probably not going to send a bunch of software engineers up there. Right? We've got robots to do break fix now. So, you know, we're good. IT is gone. So services are going to be easy. Yeah, but. Oh, I left that device back at Earth. I just think that's not going to be good. Just replicating the ones. I think there's like an eight minute. Amazon is going to have the first monopoly on next day delivery in space. They'll just have a spaceship that sends out droids to Mars. Yeah. But I think that when we start going back to the moon and we go to Mars, people are going to think, hey, I need this application now to solve this problem that I didn't anticipate having. And in science fiction, we kind of solve this with like how, right? Like you have this AI on this computer or this on the spaceship that could do all this stuff. We need that. And I haven't seen that here yet. No, it's not here yet. Right now I think they're getting the hardware right first, but we did a lot of reporting on this with the DOD and the tactile edge, you know, military applications. It's a fundamental, I won't say, it's a tech religious argument. Like, do you believe in agile real time data? Or do you believe in democratizing multi-vendor, you know, capability? I think the industry needs to sort itself out because sometimes multi-vendor, multi-cloud might not work for an application that needs this database or this application at the edge. Right. So if you're in space, the backhaul, it matters. It really does, yeah. Not a good time to go back and get that highly available data. We went highly, is it highly available or is it two terms? Highly available, which means real time and available. It really means it's on a disk. Right, yeah. So that's a big challenge. Well guys, thanks for coming on. Sharon, look for the company. What are you guys up to? How much funding do you have? How old are you? Staff, hiring? What some of the details? We're about 45 people right now. We are a globally distributed team, so we hire like from every country pretty much. We are fully remote, so if you're looking for that, hit us up. Definitely always look for engineers looking for more data scientists. We're very, very well funded as well. And yeah, so hopefully check this out. We're guys headquartered. So a lot of us live in Columbus, Ohio. That's technically HQ, but like I said, we're in pretty much every continent except Antarctica. So you're all virtual? Yeah, 100% virtual. 100%. Got it. Well congratulations and love to hear that data doc story at the time. Data robot, yeah. I mean data robot, sorry. Let's get a call to be a data company. Give me a data company. Thanks for coming on and congratulations for your success and thanks for sharing. Yeah, thanks for having us. Yeah, thanks for having us. Pleasure to be here. Mr. Cube here at Reeve R's. I'm John Furrier, host. Thanks for watching more coming back after this short break.