Nikunj Mehta is the founder and CEO of Falkonry, whose time series AI solutions are used by companies to power their digital transformation and achieve significant improvements in production uptime, quality, yield, and safety. He is the co-founder of the Industry 4.0 Club, an egalitarian effort among user advocates, technologists, and manufacturing leaders to advance industrial transformation through Industry 4.0 methods and technologies. Nikunj recently spoke with Plant Services editor in chief Thomas Wilk about the future of Falkonry, the relationship between asset management and artificial intelligence, and how detecting anomalies can improve sustainability.
PS: There's some new developments this week for Falkonry, some interesting next steps. Perhaps you could share with our listeners what's been happening this week?
NM: Yeah, so Falkonry is an almost 11 year old company, and we've just announced that we've signed a deal of merger and acquisition with a Swedish company called IFS, which stands for Industrial and Financial Systems. So that's the latest and greatest on our side.
PS: That is really exciting. IFS has been a longtime partner with Plant Services. We've gone to cover IFS World, their global conference, and it's exciting for both companies.
NM: Yeah, definitely, I think this is the first combination that I can think of where two software companies that are focused on asset heavy industries have come together to exploit the unmet potential of analysis for the purposes of improving production and manufacturing.
PS: For people who are just learning about this relationship today, let's start with what kind of customers Falkonry serves, and how you see (or Falkonry sees) this market developing going forward.
NM: This is one of the best ways to understand why this has even come together. Falkonry, since its inception, has been serving the industrial customer base. And these are multinational companies, they have manufacturing domestically in the U.S. but also overseas. And we have also been working for the last about five years now (the sixth one is in progress) with the U.S. Defense Department and the various intelligence agencies. In our beginning, we were focused on physical systems, and as you can imagine, physical systems exist in many different forms. They are certainly used in production, but they're also used for the infrastructure of our society. We continue to work with both types of customers, and it is important perhaps for your audience to know that the fact that we've worked with both kinds of customers is both benefiting from their investment in Falkonry's technology, but it also provides a long term visibility to the continued availability of our company. None of these organizations really make a choice that is ephemeral; they want to have these kinds of core operating technologies available for a really long time. And also, I would say that IFS has a very similar business where they work with both defense customers, as well as the industrial organization. So we seem to complement each other very well.
PS: The heart of our audience is centered on reliability. And not just maintenance, but focusing on understanding potential asset anomalies, streamlining operations by using the latest predictive technologies. I've had talks with some of your team before about how operations data is often kept separate from predictive data. Could you talk about how this relationship with IFS will help bring operations data and predictive data that much more closely together to help industry?
NM: A lot of operations data is produced and then almost immediately discarded or at least lost; it originates in the industrial automation world. A lot of the predictive data is used for purposes of workflow planning and workforce optimization. These two are very hard to combine because they are very different levels of abstraction. IFS is an organization that knows very well how to manage the predictive data well, and in Falkonry, there is a way to exploit all of the operations data to match it against the predictive data. One thing we have to keep in mind is that while the predictive data is helpful in guiding us in directions to take from the operations data, it has been very hard to exploit it as a matter of fact, because a lot of the details about these predictive records have been suspect, and therefore learning from it has been very hard. But what Falkonry is enabling now is learning from the operations data and mapping it to predictive data, as opposed to learning predictive data against the operating data. And in that process we've made the data quality issues, or the predictive data, a lot less ominous, and a lot more incremental in nature. We expect that with work that Falkonry is doing, we are going to contribute to the predictive data cleanup and to populating the predictive data from operations.
PS: Another question about the the deal this week: how do you see this relationship as pushing the needle either for asset management software or for artificial intelligence? Do you see both of them are informing the other and moving technology forward?
NM: You know, this is a very insightful question, how does it move AI forward? One of the big challenges that AI has had is the lack of industrializable use cases. We see a prevalence of ChatGPT discussions, and I'm sure that there will be adoption into the workplace, but save for that, most of the AI has found it really hard to get adopted. This is an indicator of the readiness of this type of AI for industrial adoption. So in that sense, it's a step forward for AI, it gives a lot of other people visibility into what makes AI successful in the workplace, and therefore encourages them to be innovative.
On the asset management side, the asset management community has been challenged with using condition monitoring characteristics in their planning process, and for the most part, they have not been able to exploit SCADA data effectively. While there are a lot of asset management software systems that will offer meters and rules on those meters, we both know that that data tends to be extremely sporadically collected, and therefore most of the important detail about why something happened is never known. That is one of the biggest reasons why people still do not exploit this asset management capability about triggers; other than schedules that the manufacturer might recommend, there's very little other use for it. This establishes a completely new mechanism that is built from the ground up on data that the operations teams already trust, with findings that explain themselves, and provide real time understanding of what's happening in the system, independently what a manufacturer might have to see. For that reason, it is enriching the asset management software world, and providing additional value to those who are using those asset management systems.
PS: I appreciate the the direction you took that answer, I can see the relationship you're describing between the two. One last question: at last year's IFS World Conference, there was substantial time spent by their teams to talk about helping companies with their reporting on sustainability and with capturing their energy use. I'm curious to know, if you see an application for Falkonry's product here when it comes to detecting anomalies? Whether you see a direct relationship there to help drive energy efficiency and reduce energy waste?
NM: Yes, we've actually explored this area a little bit in the course of understanding how we can benefit the ESG objectives of organizations. And you see reliability as an objective, you see quality as an objective. For that reason, energy efficiency as well as emissions control are also objectives of the same exact operation. We cannot achieve them with what we consider to be epidemiological data about our questions. It has to be real-time data, so that we know where, when, and why intervention is necessary, that an operations team can take action against. ESG is no different. In fact, we've had situations where our customers wanted to know when a volatile organic compound release was about to happen, that was proscribed, and therefore would be penalized. And they were going to use the same SCADA data to find out answers of that kind of question. Energy efficiency has a similar characteristic; so long as the SCADA data are relevant to any of these characteristics that we're interested in, directly or indirectly (and it's often indirectly not directly), then the same exact AI is going to be used even for those ESG objectives.