Nikunj Mehta with Falkonry/IFS

SUMMARY KEYWORDS

Falkonry, enterprise asset management, industrial, data, unknown unknowns, ai, industrial automation, process, ifs, anomaly, asset performance management, industry, people, apm, talk, require, world, behavior, realized, conversation

00:04

Welcome to the Industrial Talk Podcast with Scott MacKenzie. Scott is a passionate industry professional dedicated to transferring cutting-edge industry focused innovations and trends while highlighting the men and women who keep the world moving. So put on your hard hat, grab your work boots,

00:21

and let's go. All right, thank you very much for joining Industrial Talk. And thank you for your continued support of a platform that celebrates industry professionals all around the world. You are bold, brave, you dare greatly. You're changing lives, that you are changing the world. Industry is happening fast. And you've joined a platform that tries to keep up with all of the men and women in industry in the hot seat. His name is Nikunj Mehta. That's Ni k U N.J. Mehta. And we're talking AI that is a that is a buzzword for today. That's for doggone Sure. And how are we taking that fabulous innovation, that fabulous technology, and helping assets, industrial assets operate at their optimal condition? That's right, they're Falkonry. They're doing it, let's get cracking with the conversation. There are so much happening out in industry, so much big time. And it's happening at a pace that is so rapid I, it's, it's, I get excited just because I love the learning. I love the education, I love the ability to be able to have conversations with leaders who are just truly passionate about solving the problems, figuring out how technology and innovation can truly solve the challenges of today's market. And so here's Nick Poonch. And he he's got this product in this company called Falkonry. And it's taking AI to the next level. And how do you apply it to industry as a whole. The conversation is just amazing. He does such a great job. There's so much and I and I get this feeling inside my heart inside my mind. I get this feeling that is just the tip of the iceberg. And that goes with all of the technology and industry innovations that are taking place today. It's just I feel it's it's a feeling. It's based off of conversations I have, but it's a feeling that we're just still at the very beginning. Where is it going? That to me is just an exciting and insightful conversation. And, and Nikunj does a great job. And again, tip of the iceberg, there's so much more. I don't know what it looks like, I wish I was younger, because I don't know, it's you know, we're all timing. It's all timing. And I'm here now and I have the great opportunity to talk to all of these leaders couple of things to put on your calendar. We've got Fabtech coming up, that's going to be in Chicago, that is September 11 through the 14th. And it's a fabulous event. We're going to be broadcasting from that, that location. And I was in I was there last year, Atlanta, Georgia. And it's exciting. Once again, it's exciting. If you're not in industry, if you're not engaged in the education and collaboration, and how to leverage the innovation that exist out there. You're gonna be left behind because it's happening now. So you got to find individuals that can help you with that journey. We also have a couple of on demand webcasts out there go out to Industrial Talk.com one is specific to the utilities the digitalization of utilities, conversation with two leaders there as well as one that is aligned to the PF curve with your supply chain strategy. What does that look like? How do you overlay that so those are two on demand webcasts out there, so check them out. All right, let's get on with the conversation. Nick Poonch Mehta, Falkonry, we're talking AI, we've got the you've got to start that education, that journey and you got to find individuals that truly have your best interest at heart. Right there. No, coach, let's get cracking Nikunj Welcome to Industrial Talk. Thank you very much for your you're busy. You're a busy guy. You're uh, you're running around. You're a DC now. Thank you very much for finding time in your schedule. I really appreciate that. How are you doing?

04:58

I'm doing well. Thank you Scott for having Me on here. And I will say that it's been a terrific years and a very busy summer. So I'm glad to see you in this virtual podcast environment.

05:11

Well, you know what, you know what it is, it's the fact that I get to live my discovery channel dreams, meaning I get to learn from the best I get to just, it's, it's like, it's a dream of mine to just constantly soak up, you know, the insights that are being delivered by leaders like yourself, it's, it's pretty cool. I've just the guy that carries the six pack of beer is like, Hey, how are you? You know, very, not offensive, that's for sure. All right, for the listeners out there just real briefly, because we're going to go through a really exciting time for Falkonry. But for for the listeners, tell us a little bit about you and your, your background.

05:52

Sure, you know, I feel like I should start from my childhood because I grew up as the son of a factory worker, and specifically to a steel producer. And so I used to go in all the time to look at the control rooms and the furnaces and the mills and so on. And as I was growing up, I realized I was not going to be working in an industrial facility because I was interested in computers. And you know, fast forward, I went to USC for my computer science degree. And I later started working in computer software companies, but realized when I was at Oracle that I wanted to do something different. And so I started in the resource intensive industrial space, first at C three. And after three years, I started this company called Falcon. So my journey has been watching the industrial world from a distance, and doing everything I could in the computer world. And then eventually realizing that what I had learned and what I liked, could actually make a big impact on the industry. So that's how we just come together.

06:59

Outside of the fact that I liked the name Falkonry, it's pretty doggone cool. And I do I remember taking a tour of a steel mill. That's, that's heat intensive. That is that's major stuff. But anyway, can talk about

07:14

07:57

Just for my understanding, you're sitting there, you're in a computer, then all of a sudden you realized that there's there's something more that can be gained from the data that that's already being pulled? Is that how did you just come to that conclusion and say, I want to, I want to do this?

08:20

That's a good question. Actually. You can think about this in waves. So you remember the time when perhaps your paycheck was produced by a mainframe computer. And the payroll providers, people used to go to payroll providers so that they will get accurate paychecks and have no liability issues and so on. Well, guess what the data was coming from computers the whole time. And yet, people did not really know how to interpret this data in ways that the program calculating your paycheck did not understand that led to financial analytics and the field of business intelligence. So for a long time now, computer scientists and data specialists have known that data usually has more meaning than just the original purpose for which it is created. And what ends up happening is we don't have supplementary mechanisms to make sense of that data. And I knew this about the industrial world because before I came over to Falkonry, and C three, I was at Oracle in order to, you know, every source every form of data, and every source of data, has probably seen Oracle in its infancy. And I saw that with text and video data, and I realized that new types of data, or at least new types that attract the attention of computer software makers usually have a lot of unfinished business. And when I first came to time series data that originated the industrial world, I realized that we do not even store that data. Forget doing any interpretation of that data that told me if we cannot even store this data. Obviously, we haven't exploited the data to its fullest extent.

10:00

Yeah, it's it's a tsunami, you can always get the data you just where do you put it? I mean, it's, it's truly a physical reality, I shove it over here. But then when you shove it over there, how do you? How do you find the golden? In that pile of data? You know that? That to me? That's a tough one

10:22

man. And it's Yeah, and Scott, you come from California, and so do I, and shows, digs and pans. That is what Silicon Valley story probably is preceded by the San Francisco story. And I thought long and hard about him, it took us a really long time to figure out what is it that can help us find those nuggets? It's not an easy question to answer. And for a long time, we looked at ways of combining human knowledge with the industrial automation data. And we kept struggling because human knowledge was so imperfect. In many cases, it was wrong. And it was very hard to overcome the bias and the errors and mistakes that were creeping into that data. And so we kept getting bad results from the process. And then realize that, you know, we shouldn't just keep human opinion aside, we have to be able to extract gold directly from that DM should not think about the human knowledge as a prerequisite. And the moment we started doing that, we derived from the same original industrial automation data, very valuable insights that of course, now, people buy us fully products. But we had been struggling to make sense of in this conundrum for a long time.

11:42

That's interesting that that's not a that's not an easy pill to swallow.

11:47

I know. And we will probably have many opportunities to talk about that.

11:50

Yeah, absolutely. Because I agree with you 100%. But, but it has to be that way, because I don't I know that I'm bias. I think, yeah, I think this, this area over here is more important than this area. But then somebody else will think this, you just have it. It's just a human condition. I just, I'm biased. It just is what it is.

12:14

I'll tell you one thing. So Donald Rumsfeld used to talk about unknown unknowns. And since then it has entered the the pantheon of technical jargon that takes everybody's head for a spin. The reality is that the industrial world has so many unknown unknowns, and it has so many distinct problems, that if we focus on solving only one particular incident, we will get almost nowhere. And then we realized that there was a better way to go after unknown minutes, is when we found that we could create a lot of benefit for many industrial organizations through that route. And that is where we've been finding a lot of golden nuggets that can make a business viable to offer software to a large number of customers. And so that has been the learning over these 11 years, we've been servicing for the last year or so.

13:10

You've seen a lot. That's that's an eternity in the world of what you're in 11 years, that's an eternity because you've seen what fascinates me is the speed at which things are changing. Now, all of a sudden, you know, I use I use AI platforms all the time to come up with some sort of, you know, paragraph. I like it, and I'll adjust it, but I like it. And you've seen a lot. Take us through just the premise behind Falkonry. Today, what what are we what are we looking at?

13:51

So Falkonry has become primarily about surfacing unknown unknowns, known unknowns and unknown unknowns, if you think about this as simply knowledge, right? Knowledge is key humans are very good at. We don't have all the knowledge there is more than vino. So how do we find that which we do not know, as an example, we often start out with some manufacturer recommendations for what speed to run a new piece of equipment. But we know that we are going to find that it can be done faster, or harder than that. It takes us time to figure that kind of question out that sort of unknown unknown. We don't know what is the right speed for us to max out. Likewise, when some problem happens, and we don't know what caused it, that's an unknown if the problem was going to happen, and if it was going to happen, what would cause it? What Falkonry is become, is a software approach to find unknown unknowns and No, no, no, it's a little bit too technical, but you can simplify. No. Yeah.

14:54

So, So correct me if I'm wrong. I'm seeing a platform that sits On top of my, my operations on sort of

15:03

level two, let's say it sits on level two.

15:06

That's right. So it's looking down. And it's it's processing. And I'm, I'm hearing that I'm able to see, I'm just trying to find anomalies. Is that accurate? Yes,

15:20

that is accurate. Because that is where you first start to get a handle on what's going wrong in your plant. And this anomalies that seem to be repeating. Now there are patterns of interest. And you want to know when patterns of interest start to emerge before they become a problem. But everything originally starts out as an anomaly, then there is a recurrence and becomes a pattern. And then there is a way to find a precursor to the pattern. And that is in early morning.

15:55

Do you do you have in that, that analogy, false positives, meaning, I see that something's happened, if I perceive that something is an anomaly, I am going to elevate that anomaly to the attention of something or somebody. And then it turns out to be another point of learning where the machine just operates that way, or whatever it might be, yeah,

16:24

well, so I'll give you this answer, just like miss calibration, or sensors, going bad has not prevented us from putting sensors everywhere. False positives are not going to stop us from getting value out of them. They're either false positives are fairly well managed in a good AI environment. Because there is the wisdom of crowds, when you have lots of sensors, they don't all mislead you, unless there is a good underlying reason, they may briefly spike up, they may tell you that something is not right about it temporarily. But for them to persist for them to be severe. For there to be a group of such sensors to say something is wrong, requires that you then look into it and understand what has happened. And so we have found that while false positives can create trouble, it is one not so prevalent as a lot of people fear. And two, it doesn't cost as much effort to review something and say, doesn't make sense. I cannot do anything about it. It's okay. I understand what it is saying. And yes, it's a reason caused by for example, I have shut down I have a new product and meeting. I don't need to do something about

17:36

there's a there's a level of comfort there, that I sent it. And it's it's, again, I think you brought up a really interesting point. And that is doesn't really cost much to say, I mean, look at it. Now. That's sort of the normal way it runs, make the adjustments and and have that platform continue to be refined and learned and continue. And it does it just continues to sort of

18:01

hold on let me clarify, let me actually add more color to what you said. Yeah, it it is learning what behaviors arise. And it is learning how to predict behaviors. And in the process, it is incorporating the behavior of a normal the the anomalous behavior that persists. And through that saying, Okay, well, this is tolerated behavior. So obviously, it's no longer to be considered as a normal, it does not require somebody to provide input to say, Do not treat this as it does not require somebody to give it a name to say, if it is anomalous, what is it, it continues to observe and continues to learn about the different ranges of behaviors that can that is what it needs to do extremely well from day one. And it does not require human involvement to set it up or to maintain

18:55

it. So for me, I can I can be in my to environment. I'm out in the field, I'm out in the process, I'm out in the manufacturing line, whatever wherever I'm at, and I'm doing my stuff. I am I am operating from the perspective that I have my CMMS system I'm performing maintenance, I'm doing what I normally do what I hear you saying and correct me if I'm wrong. You're going to take that that platform itself and make it in such a way that my efforts may maintenance guy or me operations guy, process guy, whatever it might be. My moves are more efficient.

19:44

Yeah, I mean, we've been asked by our customers to provide them reports about where specific observed anomalies arise, and send it to them as a notification as quickly as is possible for the data. And that people will then look at those findings to decide what they're going to do about it. If it's an anomalous behavior, generally we want an engineer to look at it. But if it's an anomalous behavior, that has happened many times before, and you have a name for it, then what people want to know is which specific behavior occurred because there is a way to cue their action based on the name of the behavior that just took place. So both behaviors are possible, one that may require an engineer to be involved, and another that does not require an engineer and can be carried out by technician.

20:35

So that's pretty exciting stuff. I love it. I I don't I still think that we're sort of scratching the surface. You've been in the business for 10 years plus, you know, but I still think that that we're still scratching the surface. I'm going to shift gears a little bit. Falkonry has is also a part of now now. Ifs explained to us that relationship, where do you sort of growing?

21:03

Yeah, so as you know, this is very recent news. And Falkonry has signed an agreement to be acquired by ifs, it is subject to government approval, and it will take us to the end of q4. But let me tell you what it means. First of all, Falkonry has been serving in the area of production optimization, or process optimization. And as part of that, improving reliability, improving availability, and potentially improving quality. Now, all of that is seen as the realm of asset performance management. And as we know, APM is a field that many analysts have written off and have said that it's a category that came in last, it's unfortunate. On the other hand, there is enterprise asset management, and enterprise asset management companies have created sustainable long term businesses. These are companies like ifs, but many others. And for the longest time, I, as an entrepreneur have wondered what are the enterprise asset management companies doing about asset performance management. And you can see most of them have some way of doing a alert based on a rule set up on a meter, where some measure value is sent from the factory periodically to the enterprise asset management. Now, the reality is that most people have not been able to make much use of their industrial automation data with that enterprise asset management capability. And yet, companies have always said we are bringing AI, or that we are taking more of the scalar data in to help you improve your operations practices. But reality is that it's not been easy, they have not really succeeded. So ifs in their analysis judge that em and EPM have to be brought closer together. And that AI has to be the primary mechanism to enable that asset performance management. So it's a very logical combination, one that many analysts seem to feel is long overdue, and it's entirely logical for it to come together. And that's how I see it for the market. And for partners and customers, I see this as a way to provide the level of support that people are looking for to go from SCADA systems to more effective operations systems, including maintenance.

23:34

Alright, you said a lot in that. That whole conversation enterprise asset management, I didn't realize asset performance management and sort of a shift in thinking, but I think, I think because of that, you've you've clarified the the enterprise asset management approach is more holistic and global. Versus the asset performance management, which gets down into more of the macro in on that asset itself. So your as, you know,

24:12

enterprise asset management is about resources. And it aims to streamline workflows. And in doing so, we are trying to minimize inventory of parts over time P, as well as the number of people required to serve a supplier. That's what we're trying to do with an enterprise asset manager. But it doesn't really know the right triggers. And that has been the main weakness of enterprise asset management. It doesn't know when it's dependent on manufacturers, it's dependent on tribal knowledge is dependent on break fix, but not really, what is the machine saying? It doesn't really have a way of knowing that asset performance management ought to have answered that question, but it So that's why everybody's been talking about smart benefaction, because APM has been around for 20 years. Why did we need smart manufacturing if APM was effective, so clearly, it hasn't done its job. Now, when it comes to APM, a lot of the APM techniques of the past, we're looking at some downsample rate of information from some key parameters of SCADA. And using that to make decisions, and what we're finding is that that's not enough, when the industrial automation system can run 100 values per second in maintaining the color, the performance of the production process, but you're looking at it maybe once every five or 15 minutes, then you're missing all the behavior that happens, that will tell you why something went wrong. So therefore, a PMS cannot answer the question, or cannot help you answer the question. Why did it happen? And without being able to answer the question, why did it happen? It's all values in question. Yeah. And that's where AI is being asked to step up. And not only answer the question, is something bad about to happen? But answer the question, why does the aI think something bad is about and therefore guide people to solving those problems, so they can avoid those problems?

26:20

This is really an interesting conversation, I have one question, do you find let's just be real. manufacturers want to manufacture process, people want to process, you know, do whatever is necessary for process and so on. You're introducing something that is not a part of my thinking, quite frankly, I, I've, I've done this for so many. But now all of a sudden, I've got to rethink and change the way I look at my my manufacturing. Yeah. How do you how do you? How do you approach me?

27:08

Yeah, I sympathize with you, I understand the difficulty that is inherent in this. Now, that's one of the big challenges with any field that is undergoing such dramatic change. First of all, nobody has the role description, to do what this AI might ask it to do, which is set it up or interpret results or reason about why it did what it did. Nobody has that role. So it's never been. But it's not going to happen while the rule is not there. But even otherwise, I think ultimately, this is a top down and a bottom up experience in organizations a top down is I need to invest in fun, these kinds of big changes. Because it makes such a huge difference to our competitive ability. The bottom up is, I already have my hands full, I have a huge quality of life issue already without having to go into dirty hazardous places in the middle of the night, with when my brain is scrambled and I have to get production back up because otherwise millions of dollars are at risk. So there are quality of life issues for the bottom up side. And they are they are not ignorant, they don't want to be roadblocks, but they have knowledge and they're on the ground presence needs to be exploited in a way that they are able to contribute to the improvement process. We've seen this from both sides. And so what we've learned is that initially people have to create, people have to be able to trust this kind of AI. Until that happens, adoption is not possible. And that trust requires bottom up experience. It requires studies it requires providing people with examples from their own world in ways that they can understand so that they know now if I'm going to let the system do some of this work, that I can always understand what it has done. Once that is available, then people are able to sidestep from the constant need to manage an AI system to simply say, Okay, give me the results of the year and will act upon. But remember that stage cannot arise until there is trust. Unfortunately, a lot of times, vendors were asked to go to that stage of providing alerts before trust had been developed. In the process, a lot of time got lost. And people may have felt that the CXOs were imposing technology that was not ready for them yet. But what we are seeing now is that there's a lot of interest in the adoption from process engineers and reliability engineers and maintenance supervisors, because they know that this is inherently going to improve their quality of life. And what they want to know is can be trusted And once they are able to trust it, the change management process is, is as simple as previously, they used to get told by vibration analysts that this particular machinery needs this intervention. And here's why. And that realize that vibration analyst could be a contractor, it doesn't have didn't even need to be an employee, it could be a service provider, right? In the same way, people are looking at AI as also a service and an AI provider as a service provider, who just gives them the alerts, they don't have to understand anything else. And those alerts are also in the same data that they eventually go to look at when there is a problem to try and figure out why. So it's no different. Except here, it is telling them to anticipate that trouble. And if they did not take action, it tells them bite my

30:51

eye out of all that great content or that information in that position. Trust. Trust is something that that is so powerful, and you're right. In a past life, I dealt with trust from a data collection perspective, do I trust the data? Trust? And and if I did, then life is sort of, okay, cool. I can move on. But when it's not there, that's, that's really good. Well, we're gonna have to wrap Yeah. Sorry. Great conversation. I could talk for hours. I don't I'm not sure if the listeners want to hear it for hours. But, but I can I love it. Yes. How does somebody get a hold of you? Now that you've gone through this ifs, Falkonry all of the changes, but they want to talk to you? How do they get a hold of you.

31:45

So you know, I do try to reach out to people through the Falkonry newsletter, I write most of the content myself. I also have made myself available through groups like the industry for our club. And while I may not have participated, let's say the last couple months, I've done a lot of work with them for a couple of years. But more interestingly, I think now this is going to give me an opportunity to really mingle with customers, as well as users to help them overcome some of the challenges they see in their own journeys. And I'm happy to sort of make myself available, you can reach me, nikunj@Falkonry.com. Hopefully, that address will still be accessible to people. You can do it through LinkedIn as well. I'm always like,

32:32

I'm a big LinkedIn fan. Absolutely. Struggling to fan. And it's sort of like, Alright, cool. Then I get to look at your profile and all that stuff. You are absolutely fantastic. I really enjoyed this conversation. It's exciting. From my perspective. Again, I think it's tip of the iceberg. There's so much more. I'm glad people are like you are dealing with it just because I think it's an exciting time.

32:55

Yeah, thank you appreciate the opportunity to talk. And I understand that this is a technically very involved area, it will get easier. I think it's already become a whole lot easier. We're not talking about algorithms and methods. We're talking about people and trust that tells you how there has been progress.

33:11

Now that's major progress. And I think that as as you continue to evolve and progress, it's going to be more powerful, more understandable. Because I think that everybody pursues that that user Enos I guess that's that's important. All right, you were absolutely wonderful. Thanks. All right, we're gonna have all the contact information out on Industrial Talk.com. So if you're not reach out to No, Nikunj, did I get that right, Nikunj? Yep. Got it. All right. Stay tuned, we will be right back.

33:49

You're listening to the Industrial Talk Podcast Network.

34:00

I'm excited. I'm excited about the future of industry. The reason I'm excited about the future of the industry is because you have leaders like Nikunj you have companies like Falkonry. Just blazing the trail, solving the challenges of today, in industry. It's a bright future. We have to succeed. The way I only know how to succeed succeed is to constantly educate because it's happening at a blistering pace. And how to collaborate, collaborate with leaders so that you're not left behind. That's what's important here. That's what's at stake. So you got to reach out to the coach all the contact information for him is all right out on Industrial Talk. We have webcasts, we have other podcasts, go out to Industrial Talk, find out more reach out to all those people because they want to collaborate. People be brave dare greatly don't hesitate reach out to Nikunj and you're going to change the world