3. IntegratedML in InterSystems IRIS (Thomas Dyar)

Speaker 0 00:01 Welcome to data points, a podcast by InterSystems learning services. Make sure to subscribe to the podcast on your favorite podcast app such as Spotify, Apple podcasts, Google play or Stitcher. You can do this by searching for data points and hitting that subscribe button. My name is Derek Robinson and on today's episode I'll chat with Thomas Dyer, one of the product specialists at inner systems about integrated ML in InterSystems, Iris Speaker 1 00:26 <inaudible>. Welcome Speaker 0 00:39 to episode three of data points by InterSystems learning Speaker 2 00:42 services. My name is Derek Robinson. As I've been mentioning in our first few episodes, we're excited about the launch of this podcast and we've already released two other episodes along with this one for you to check out. In this episode I'll be talking about machine learning and integrated ML with Thomas Dyer. Thomas is a product specialist here at InterSystems, focused on the area of machine learning. In the interview, we're going to start with some interesting perspective on machine learning in general and then segue into one of the exciting new features that's coming to InterSystems IRS data platform. And that's integrated ML integrated ML is really built for the SQL developer who wants to incorporate machine learning into their application, but they may not have the resources at their disposal to do it the traditional way. I'll leave the more thorough explanation to the expert. So here's my interview with Tom. All right, and welcome to the podcast, Thomas Dyer, one of our product specialists here at InterSystems. Thomas, how's it going? Speaker 4 01:37 Very good. Thanks Derek. Speaker 2 01:38 Yeah, so, uh, today we're going to be talking about machine learning, which is one of your areas of expertise both in your career, I think, and here at InterSystems with your product and kind of the products that you oversee. Um, so let's dive right in and get started. But before we get into the product, let's talk about machine learning in general. It's really a very popular new topic that's been emerging in the last few years as something that's really, really relevant. Um, what is machine learning to you and how did you first get interested in it? So machine Speaker 4 02:05 learning is essentially taking whatever data that you have and asking a computer to figure out what it is about that data that's interesting. And that also could be relevant to what you want to do with your application. Right? So I got interested in machine learning in high school and really what I was interested in at that time was how the brain worked and being able to model the brain in a way that you could take the, the crazy complexity of all the neurons and distill that down into a mathematical representation of how those neurons would work and make it possible for a human to think was just extremely interesting. And as I got into actually trying to build those kinds of models, uh, I realized that there were a lot of applications that you could, you could actually build that would be useful to humans. Speaker 2 03:05 Yeah. Interesting. So, uh, you know, to be honest, I hadn't actually taken that perspective on it, but it's, it's kinda cool to frame it that way. Um, so like as you've gotten more into it, what are some of the most important and kind of significant maybe life changing applications that you see in the world today that involve machine learning? Speaker 4 03:21 So I'd say one of the most exciting areas is healthcare. Healthcare is a, is a realm where we've spent a lot of time last 10, 20 years taking all of the data from, from doctors and putting it into computers. Now you have a mountain of data and you have extremely life altering problems to apply that to whether you want to learn how to treat someone better or avoid some kind of a, a bad outcome, say a complication from surgery, you now have a huge amount of data that you could look at and decide whether or not, uh, what you should do with that. Right, right. Speaker 2 04:06 So, um, so yeah, so I think health care is really a well known one. I think with, um, machine learning. And I think what you brought up there is, is really important, not just from how well suited it can be for that amount of data, but also how significant it is in, in that it has impact on lives. Um, any non-healthcare examples come to mind that really jump out at you as good use cases? Kind of cause I think for, in some of our previous episodes we've talked about how sometimes people see inner systems and they associate healthcare it cause they see our logo and they see how big of a player we are in healthcare. Um, but for the people that aren't in healthcare, there's a lot that they can leverage on our technology stack as well. Um, any, any good ones come to mind as far as really cool machine learning applications that aren't in the healthcare space that you've seen? Speaker 4 04:48 So definitely in the area of logistics, uh, is a, is another, another application field where inner systems technology is used. And there you have, uh, kind of optimization problems. Things like, well you have a shipping company that needs to optimize how they move their, their ships and their infrastructure around to be able to meet the customer demands in real time, be able to, uh, lower the costs there. Machine learning is also very applicable. You can throw a bunch of data at it and determine the best place to, uh, put a ship or put a, put a box and be able to, um, to get it to the customer and the optimal boy. Speaker 2 05:33 Right, right. Interesting. Yeah. And so that kind of, the more you think about them, the more it opens your mind to even more examples that can really take that data and use it in an effective way. Um, so let's transition this into our inner system stack, right? Which the, one of the new features that's coming out for InterSystems, Iris is integrated ML. Uh, that's one of the products we mentioned that you are managing and overseeing here at inner systems. Um, tell us what integrated is kinda who it's for and what it can do for those developers. Speaker 4 06:01 So integrated ML is a new capability that's that we're placing within our sequel environment. So it's an all sequel feature that is turnkey machine learning. So don't need to install anything. It's going to become out of the box, uh, able to be applied to your problems. And then it's going to bring the best of breed machine learning frameworks right into your sequel environment. So an application developer that knows the data that knows sequel is going to be able to train models and then be able to use those models to make predictions and make their application smarter. Speaker 2 06:41 Interesting. Okay. So that, that's a cool kind of a brief explanation of what it does. And I think in the little amount that I've kind of been looking at this stuff and kind of seeing it from the curious perspective, looking at it more closely, should, should these developers think that it's just instant magic, right? Let like is it, is it just kind of like flip a switch and all of a sudden you don't need your data scientists anymore? Um, or is it more of a thing to kind of open the door for them? And kind of get them along the right path, let them taste it a little bit and kind of gets you prime to be able to go even further later. Well, how would you frame it for those developers? Yeah, it's definitely not Speaker 4 07:14 to replace data scientists that there's a lot of heavy duty statistics and math and, and also judgment that's required in determining where machine learning is useful and not. But um, as far as the ability to get a read on whether or not your data is good for machine learning, integrated ML is a, is a, is a great place to start and we'll get you the basics of a machine learning model and allow and also focus on being able to use the predictions from a machine learning model right in right in an application. So instead of spending all your time trying to install some framework, learn Python, figure out how to go between Python and sequel and all those things that are necessary to actually put a machine learning model into production. Integrated ML is just, it provides all the plumbing and makes that, that process a lot easier. So you can focus on your data, focus on the problem and, and let the, let the computer do what it's good at, which is, which is uh, putting it all together. Speaker 2 08:24 Right, right. Interesting. So, so it kinda, it sounds like you know it for the developer, like you said, that knows their data, that knows how to use sequel, that knows how to re is really focused on those things in their application. This lets them take this buzzword that they've been hearing all over all over the industry, right. Machine learning and being able to apply some models and create some models and train those and really effectively use them on their data. So for some of the audience that might be a little bit more, have a little bit more expertise in machine and maybe on inner systems, IRS data platform as well as maybe legacy products have used the previously existing architecture, which could be the spark connector and loading PMML files for your machine learning models. How does that kind of play together with the new vision of integrated ML and InterSystems Iris? And what would be the, what would be the message to the people that do have some experience with machine learning in that context and now seeing integrated ML come into the picture? Speaker 4 09:21 So the, the, the focus of integrated ML is really on, on the sequel experience and the SQL developer, uh, so it is really there to kind of streamline the process and also bring these frameworks in, whereas bring these frameworks into the development process and make that really easy focus on deployment. What we've had before is more of a, of a, of a feature set for experts. Uh, so our spark connector was good for, uh, users that all already had either a spark cluster or experience with spark and wanted to then take that, take the, the models that they would build in that environment and easily deploy those. Whereas integrated ML, you don't need to start out with a spark cluster. You don't have to have a huge infrastructure there. It's going to, to provide that OnRamp to machine learning much, much more quickly. Speaker 2 10:18 Right. Kind of a, an easier way for, for these developers to access this and remove removing barriers that, you know, previously existed to really implement machine learning into their applications. Exactly. That's really cool. So kind of wrapping up in summarizing a little bit, what do you see as the future of integrated ML now? I know you've kind of given us a, an example of it. Um, when can developers maybe look to hopefully try to see more coming out about this and like anything you kind of see as the, as the vision and like what if someone right now listening to this is interested in what you're saying and kind of really, it's kind of intriguing them as an SQL developer. What's their next step and kind of what, what should they look for in the future? Speaker 4 10:53 Uh, well it's coming out, uh, in the spring and anybody that's interested can get on the developer community and look for content there. Ask questions there. And if you don't find what you're looking for, feel free to email me. My email is going to be in the description of this podcast and I'd love to hear from you and find out, uh, any feedback you have and any questions I'll try to answer. Speaker 2 11:16 Absolutely. So, Tom Dyer, thank you so much for joining us and we'll see you next time. Thank you very much, Derek. This was fun. Thanks again to Tom for sitting down with us. I really liked this perspective at the beginning on what got him into machine learning and kind of how powerful it can be. It brings into context just how amazing thinking and the human brain really is and how it all works and really highlights how impactful some of the advances in technology can be. That can bring us closer to replicating that magic on the integrated ML front. There'll be much more content in the coming months for you to check out. As Tom mentioned this spring, it'll be released for general availability with InterSystems Iris. In the interim, if you're super eager to learn more right away, shoot Tom an email. We put his email in the description of this podcast, but otherwise, definitely leverage the developer community, community.intersystems.com and browse our learning [email protected] to learn more about machine learning features and InterSystems Iris features and kind of how it all can fit together and kind of pique your imagination of what you might do with it down the road. Speaker 2 12:18 I'm sure there are users on the developer community that would love to chime in with their thoughts as well. So hopefully you enjoyed episode three and remember, make sure to find us on your favorite podcast app and hit that subscribe button. Thanks for listening and we'll see you next time on data points.