Join the Insider! Subscribe today to receive our weekly insights
David (00:00)
Welcome Shiv Trisal. Shiv is the global manufacturing, transportation and energy go-to welcome-market Leader at Databricks, a really cool Data and AI company. transportation and energy go-to a career across firms like like Ernst & Young, PWC, Raytheon Technologies, just to name a few, Shiv is reshaping industries with data and AI. His expertise and experience uniquely positions him to discuss the transformative power of data in industrial digital transformation.
Thanks for joining this call.
Shiv Trisal (00:30)
Thank you, David. Thanks for having me.
David (00:33)
So, Shiv, let’s dive directly into today’s topic. So you’re driving the worldwide adoption of Databricks within these sectors, so manufacturing, transportation, energy, and so on. So first of all, for those who don’t know Databricks yet, what would be your elevator pitch?
Shiv Trisal (00:54)
Yeah, so I mean, think about it like we’re all surrounded by generative AI now in the world, right? So in the world of generative AI, you can have the general intelligence that the internet has to offer and things tools like chat GPT have made
But what enterprises want is not just general intelligence, they want specific intelligence on their data that’s built on top of their enterprise data.
and the reservoir of knowledge that is OT information. And that’s what we help companies deliver. We help them deliver what we call data intelligence, which is intelligence that’s specific to their organization and that helps them democratize gendered AI across the enterprise.
David (01:36)
Yeah, and I think the elevator is now on the next floor. So that’s perfect timing. But Do you see a difference between working with IT data and working with OT data?
Shiv Trisal (01:40)
Ha ha ha.
One of the beauties of Databricks is all data is treated the same way. So whether it’s structured or unstructured, whether it’s things like tables that are coming out neatly in rows and columns from your enterprise applications, or it’s highly unstructured sensor data, or you think about natural language information, things that people key in on the shop floor or near the asset in the field service, all of that is treated the same way in Databricks. You don’t need a bunch of tools to handle unstructured data and structured data.
handle it in the same way and provide the same experience to people who are trying to interact with these data sets. Regardless of what type of data they’re interacting with, their experience remains the same, which is why it’s so powerful.
David (02:31)
Yeah. Now you’ve been at Databricks for two years. Before that you were, I would say, on the industry side. Same for me. Two years ago I was also on the industrial side. Did you see the way data is used, the way people think about data? Did you see that changing over these years when you were in the industry?
Shiv Trisal (02:58)
I think so I come from the aviation industry, you know, there was a market change in realizing that the market’s only gonna grow so fast, right? So you’re gonna have build aircraft at a certain rate and the market’s expectation is you know You have to grow beyond the platform You have to grow beyond that growth rate and the way to do that is be better at
using data from the entire life cycle of the product and really drive better decisions in how you service the fleet.
operated in how you design the next generation of technology that’s going to go on airplanes in the future. So, you know, that’s really what the realization that kind of snuck in somewhere, I would say early 2010s, mid 2010s. And I worked with a number of customers in my consulting career, advising on that and being on the forefront of it and then got to do it myself in a more involved capacity at Collins Aerospace, which is part of Raytheon. And I think that’s the same motion.
in what the deepest depths of manufacturing and energy, specifically in the industrial automation space. Not to say that the industrial automation space has not seen high degrees of automation. I think there has been market improvements, but the value of AI and machine learning and now even generative AI is to learn and provide insight into patterns that humans cannot see. And I think that is very, very valuable.
especially at a time when
when manufacturing companies and energy companies are facing with this challenge of having an aging workforce and they really need to codify, what are some of the main things that the new talent has to account for and learn really, really quickly and ramp up really, really fast because we can’t hire people fast enough and that’s the biggest risk that’s being faced by these companies today. And that’s really where the capabilities in generative AI
big role to play is to codify, use this vast data that you have, the enterprise and OT data that you’ve harnessed for so many years, use it to train the next generation of workers to help them be more productive in 30 days, as opposed to waiting on 30 years, which we don’t have, to get them to the same level of proficiency as some of the experts today.
David (05:26)
When you were talking about the aviation industry, in typical manufacturing, we talk about green field plants and brown field plants. Where I would say most of the projects we do are in brown field plants. We use the sensors which are available. Sometimes we install a couple of new ones. We install some IoT. We have this spaghetti type of approach. No, approach network.
Shiv Trisal (05:34)
Yep.
David (05:52)
different protocols, all these things. When we built greenfield plants, I would say we typically try to implement some newer technologies, some more standardized ways of working. Were there similar things in the airplane industry? Do you have like greenfield planes and brownfield planes as well?
Shiv Trisal (06:16)
Yeah, so yeah, happy to give you a sense. Like, I’ll bring it back to the experience back into the industrial space for, but yes, to answer your question, there were newer generation aircraft that generated way more data, and then there were the older generation aircraft. So just to compare it, to give you a sense, right? The seventh, let’s take the A320 example, right? So the A320 Classic, which is the most popular airplane on the planet, right? I’m sure many of you fly it all the time from Airbus.
You know that had like a hundred to four hundred parameters that you could offload from the aircraft right like that was Pretty much what it was
But a newer generation aircraft like an A350 or an A320neo would have like 1,700, 1,800 parameters that you could offload and you could configure even more because there were just better routing technologies, better connectivity, better sensors on the aircraft that you could take advantage of. And then you also had the networks to go move this data off of the aircraft and bring it into analysis platforms. So that’s sort of in a nutshell, but how it connects back to the world.
of sort of industrial data is fundamentally, you know, Companies realize that, you know, the unit, like for example, the asset or the production equipment or the line, you know, it’s just one, you know, one sort of node in a connected ecosystem, right? And from my perspective, you know,
the companies that I have seen do really, really well in data-driven transformation in this industry are those who can ask us like, hey, now that I have a very crisp answer to this question, which is now that I have all of this data from my shop floor equipment or my assets in the field, like what decision am I driving? What decision am I driving better today with all the data that I have versus, or what am I trying to get better at in the future?
a very clear answer about that, right, versus getting mired in these unnecessary acronyms and you know debates about
you know, which architectural paradigm in the plant side is the right one. I mean, those are all very important debates to have, but the more important question is, now that you have all the data from all of your sites and all of your assets, what decisions are you gonna inform? What decisions are you gonna make better? So that’s really what, you know, comes back to it. In the aviation industry, in my experience, the question was around, hey, can I, you know, if I predict a maintenance event,
ready to catch up with the airplane in the next location. That was what we were trying to get better at. In industrial systems, and the way I see the world, is you think about your production layer and how it interacts with things like your supply chain. How does it interact with, hey, how much, can I, do I really have the capacity to ship things?
at the rate that I’m producing or am I just shipping air? Do I have the capacity to think about, hey, if there are demand spikes or changes in demand, like how, none of that information is useful unless I can make a decision on it, in which case it would be more about if there is a change in demand or fluctuation in demand, can I take advantage of it by revising or being more flexible in how I schedule production?
also understanding, let’s say, you know, if I have a production gap today, you know, where is the Delta and how can I make up for it tomorrow? And that’s really, those are the kinds of decisions that, you know, I feel like some, the top decile performers in our, in our business. And we have about 1800 customers that we work with in this industry is that’s the difference that I see between the, the people who are really good versus the people who are.
David (09:56)
Yeah.
Shiv Trisal (10:16)
kind of just juggling technical acronyms.
David (10:20)
So you have all these customers and here’s just a story from my side. So one of the reasons why I started writing the IT/OT, insider was because I’m always talking about data, data platform, data architectures, AI, and those things, that’s my thing. So I’m going to a meeting and I would say, here is my typical meeting.
left side of the table, OT guys, right side of the table, IT guys, me sitting somewhere in the middle, trying to be the translator going left from left to right and right to left. Now the interesting thing is that I hear from the, I would say the more IT side, data science, data engineering side, Databricks is mentioned in literally every meeting, I mean. So that means that
I think you’re doing a really, really good job there. But what I see is I see quite a lot of translation issues where for some reason, I would say on the IT data science, data engineering side, what I see as a translation issue is that a lot of people, they think that by just applying mathematics, by just applying data science principles that they, I would say can solve a physical problem.
Shiv Trisal (11:19)
Thank you.
Yeah.
David (11:46)
I’m a mathematical engineer myself, so I’m allowed to say those things. On the OT side, you see, what’s the right words? A big gap, people going like, who are these people? Are we now just supposed to be the ones who send data up and then it’s up to them? So
Shiv Trisal (12:03)
Yeah.
David (12:13)
you see a bit of a divergence between those two groups. Well, at least in my perspective. And I was wondering, were you in similar meetings? Do you have similar experiences? And how do you try to, I would say, how do you explain to them that it’s not one or the other, it is the combination of both. You need a strong OT foundation and you need a very strong IT focused data platform
and data capabilities. So were you in those situations yourself?
Shiv Trisal (12:50)
Yeah, I mean, see this quite, I would say, I would say I’m seeing it less and less, which is a good thing, right? You always have to take stock of like trajectory, right? Like, cause you can talk about it, oh, IT and OT are two separate worlds and like, yeah, but they’re moving closer to each other. And I don’t see it as one is gonna dissolve the other, right, or kind of absorb the other. I just don’t see that happening. But at the same time, like, you know,
complementary technologies that have to work together to deliver the outcomes that we were talking about earlier. Like the priorities of the business are changing, the technology impact on the business, and the speed at which things move now are just astronomically different than ever before. So the worlds have to come together. So what I see a lot of times is, and I think what’s important to acknowledge also,
before we talk about this is There is no substitute for domain expertise. So let me just start there, right? There is no substitute for that. No amount of data without domain expertise and modeling and knowledge is useful. And so we have to account for that. And any OT, IT, these kinds of conversations where you are in the middle, David, like that’s the first point to start with, because if you ignore that, you’re pretty much destined to fail regardless of which side of the table you are on.
And the way to think about it is, you kind of have to deliver two or three things here. One, you have to think about how do I get data from these OT and edge systems and to be able to combine it with enterprise data at scale.
How do I do this not just at one site or one asset or one plant? How do I do this across thousands of sites and thousands of plants and really discern insights at an enterprise level so that you can learn from what works and replicate it at scale? And this idea of thinking about this is not very different from how most manufacturing energy companies have management systems. They have safety management systems.
I personally think now they should just have a pillar for data management systems because you can’t report on safety, quality, reliability, anything at an enterprise level. If you don’t have your data in order, how do you know what you’re reporting is right? And how do you know whether what you are doing is actually working? So I think to me data is such an important pillar now. It used to be technology was sheepishly coming into a pillar of our management systems inside of these organizations.
But I feel like tech is important because you need to you need to connect your shop floor systems and all that stuff But at the same time, you know data is becoming its own pillar And and I think people process culture all very important, but data is an additional Pillar that you I think most companies are raking up to that and I do see that changing Over over the next few years and it’s happening as we are speaking right now And in a lot of the organizations that I speak with that is very
intentional to say, hey, data. So we work a lot with Mercedes-Benz, and data is such a big pillar of their MO360 strategy, which is their Mercedes-OPS 360 platform. They produce 2 million units a year. And fundamentally, data is such a big pillar of that. And they’re trying to analyze data for basically creating a digital twin of their entire vehicle production process and understanding, hey, it’s not just production issues that can impact my performance. It can be supplier issues.
be logistics issues. Knowing how interconnected my ecosystem is and the digital twin reflects that I am much faster to actually go identify these issues and actually proactively do something about it so that the plan that I had for that day, for that month, for that quarter becomes more and more certain over time and that’s very important.
David (16:52)
The more I talk and think about, I would say, the term IT/OT convergence, the more I’m convinced that the only thing we’re actually talking about is data convergence. Because as you rightfully said, I don’t really see…
Shiv Trisal (17:03)
Yes.
David (17:08)
I would say the pure OT work, the pure OT scope changing. We still want to have the real time control and we still want to have the execution on shop floor level, whatever shop floor might mean, but whether you have a, I would say, a more traditional manufacturing plant or you’re in the energy industry or you’re in critical infra or whatever, I would say you always have that same type of things you wanna do real time safely and as optimal as possible on the shop floor. So,
Shiv Trisal (17:12)
Mm-hmm.
David (17:37)
I’m seeing IT / OT convergence more as data convergence and data convergence, I think, is only about one thing. It’s about scaling data-driven initiatives, making it really, really easy for people and applications, well, typically applications, people will probably access that data through an application, but to make it as easy as possible for people and applications to have the right data available at their fingertips.
Shiv Trisal (17:41)
Yeah.
David (18:06)
without any hassle without the need to understand, I would say, the asset structures and context layers and without the need to manually integrate several data sources and so on. Because once you’re able to scale your data applications, then suddenly you allow this, yeah, you allow the, I would say the real why about data to happen. So I think we’re moving away from IT/OT convergence as,
Shiv Trisal (18:07)
Yeah.
David (18:36)
Yeah, it’s us against them to this data convergence thing. And then the scalability becomes really important. And then I would say, my question to you is, do you also see that happening? Do you see these examples of scalability happening in the world where it becomes easier and easier to try and build something?
Shiv Trisal (19:02)
Yeah, I think you put it really well. We see the world very similarly, David. I don’t know, maybe we are lost cousins from a previous thing. But fundamentally, what I see is like, if you look at the world of IT and OT, and I personally, the convergence thing is an easy term to use. But fundamentally, the goal here is…
You know, OT people and data people should work off of like when I see data people, I mean data scientists and the next generation of sort of AI technologists.
they need to work off the same set of data, fundamentally. If you have to make industrial AI successful, they have to work off of the same set of data, and they have to manage the quality of the information that they’re training these models on. So if you break that and you go rogue, which some data team members do sometimes,
I think it’s good to do a bunch of POCs and pilot projects, but invariably you’ll come back to the question of does it scale? Does it scale across the, because the point is not whether you can do it for one line, the point is can you do this across the entire production ecosystem?
So to me, I think that’s the most important realization is that the IT and OT need to work off of the same set of data. And there has to be a stringent focus, a continued focus on data quality and the ability to then take insights that these new AI technologists are working on, right? So your data scientists, your machine learning engineers, your generative AI people.
Like all these insights can’t live in that world. They have to go back into the OT world because that’s where the decisions are made. So those three things are so important is to get that data out, work off of the same set of data, go into this aspect of how do you manage the quality of it over time, and then be able to share the results of your work to decision makers so that they can end up making better decisions.
And to me, that is the promise of data convergence. We announced a partnership on these lines with AVEVA last week, basically identifying that as the pain point that we are solving for. Can communities work off the same set of data, even though they may have different ways of building things, and then making sure that the quality of data improves over time based on everything that they’re learning from their operations.
And then once they’ve built these next generation solutions, data applications, or even machine learning applications, pushing that back into where the decisions are made. And sometimes it’s on the edge. Sometimes it’s on web applications. All of those deployment options are found. So the way we think about that is if you simplify and consolidate the core, you can drive more optionality in how you choose to serve the many different stakeholders that you do need to serve.
in this
David (22:10)
It definitely resonates to me, and that’s exactly where I think AVEVA plays a really, really important role here because you can find them in the entire stack. They have a, I would say…
They have a very broad portfolio in the operate, the executes part of the OT world. They have with their different historians also still an on-prem footprint. Then we bridged the gap from on-prem, maybe a bit of edge computing to the cloud. A couple of years ago, I would have said that cloud adoption is still…
rather slow in manufacturing companies. There was a lot of hesitation, people going like, yeah, no, no cloud for us. We’re a physical company, right? So we need those physical servers. I see that changing quite rapidly. Do you?
Shiv Trisal (22:56)
Mm-hmm.
Yeah.
Right.
I mean, I lived through it.
David, like, so I was in the aviation industry when the pandemic hit, right? So think about that, right, for a second. This is an industry that literally overnight went to zero, right? And the only way you could talk to your customers was through digital means. So I think the industry learned a lot from having to, and not just like, I don’t mean just the flying public customers, I mean your suppliers, your, you know, everybody back in the ecosystem and just percolate.
David (23:28)
Yeah.
Shiv Trisal (23:45)
all the way back.
I think the rates of digitization and cloud adoption and the need to be interoperable as a result just took off. It was the moment that the industry, I’m sure there’ll be a study done five years later that looks at this and comes back, because I lived through it. I have a unique perspective. It’s like, yeah, like this, that was the moment that everybody realized like, this is it. Like, if we don’t invest in the ability to have interoperable systems
ability to really digitize our processes and really focus on data, we’ll have no business. And I think that’s where the tide really turned. Nowadays, it’s less about like, I think the question I often get is, well, I have a plant level historian, and it’s on-prem, and I’m happy with it. What is this cloud stuff you’re talking about?
in an organization, there’s always a digital transformation or a data chief data officer that’s wanting to sort of monetize the company’s data that’s their sort of, you know, remake. And then people are like, well, you know, I like my system, I like my OT system, you know, it’s on the plant, like everything, which makes sense, right? I mean, it’s doing the job, you know, so there is definitely, it’s super important. That’s why companies like
users and those user bases. But at the same time, you know, if you’re trying to get more predictive and you’re trying to do these things at scale, if you’re trying to do this across plants, across sites, in a way that you can replicate your successes, you need a broader enterprise approach. And in the data space, Databricks brings that, right? We don’t do this for every layer of the stack.
But where it truly matters is data powers everything. And we do it, we provide that ability.
to be consistent in how you ingest data, how you curate data, and how you serve it up, you know, provide flexibility in how you serve it up for different stakeholders. And super important given the trajectory of the industry. And I think cloud is becoming less and less of an issue in adoption that I’ve heard, like, let’s say comparing it to like five or 10 years ago.
David (26:16)
It’s an observation I share. I think also for me then, for my background in the chemical industry and also now more and more in the food and bath industries. Actually, during the pandemic, I even set up my own rogue video server just to be able to stay in touch with my team. Because at that point in time, and especially in the first months of the pandemic,
we didn’t have the means to start from just, I don’t know, hundreds of concurrent calls to go to thousands or 10,000 of concurrent video calls. We didn’t have the means to do that. So I set up that rogue video server, which I didn’t dare talking about because I thought like, IT is going to kill me. But then indeed that started.
Shiv Trisal (26:54)
Yeah.
Yeah.
Hahaha
David (27:13)
I would say that the mindset started changing so rapidly. The more I look back, the more I think that the actual industry 4.0 era has only started since the pandemic.
Shiv Trisal (27:18)
Yeah.
David (27:33)
it’s before that we were doing small proof of concepts
Shiv Trisal (27:36)
I think… Yeah.
I think it wouldn’t be too much of a stretch to say that in my opinion. And what followed after the pandemic, obviously the health, public health issue was a bunch of supply chain disruptions and people are like, whoa, things that used to happen once in a decade now happen once in a year. And maybe two times a year, maybe three times a year as we’ve seen since. So it’s been a good almost three years or four years now
the original sort of outbreak. And since then, we’ve seen how many supply chain disruptions, right? Like things that used to happen once in seven or maybe in a decade, would happen like once or twice a year now. And so what that fundamentally means is, apart from the public health issue driving digitization, I think what fundamentally that means in my opinion is, whatever you thought was fast today is slow tomorrow.
So if you thought you were making decisions at a certain pace and that’s keeping up with the market, you’re wrong. Because these things, the ecosystem around you, your supply chain, where you source materials from, the suppliers and logistics companies that are serving you and getting you inbound logistics and outbound logistics services, all of that is susceptible to what’s happening in the world.
can move is be much better at how you leverage data and drive these predictive capabilities and having the ability to manage by exception. Otherwise, you’re running blind for the most part. If you don’t have control over this, I would argue you’re really running blind in your operation.
David (29:32)
I was looking for a way to end this call, but I think you perfectly summarized it. We could go on for hours and hours and hours, but until I would say until the next time we meet, thank you very, very much for joining this. It was really insightful. And I would say, yeah, good luck with your endeavor at Databricks. Thank you very much.
Shiv Trisal (29:46)
Yes.
Thank you, David. Thanks for having me on. And hopefully, we do this sooner again than later. Thank you. Bye bye.
David (30:03)
Definitely. Thanks.