Hello, everybody, and welcome to the episode of Espresso 4 .0. Today with us, we have a very special guest all the way from Brazil, Adriano Costa. We will be talking about challenges of connecting and implementing AI and all kinds of challenges that companies might have with integration of different systems, because Adriano is a formal specialist on that. But without me actually telling you about Adriano, I'm going to let Adriano Costa introduce himself, his background and what drives him in his business.
Adriano Costa (00:34)
OK, hello, Filip and Wizata team. It's great to be here. Thank you very much for the invitation. My name is Adriano Costa. I'm an Inovex director and I have been working with industrial software solutions since 2005, and specifically with connected software since 2012. In 2020 we opened Inovex to help professionals and their companies to look and find solutions to connect data from equipment, industrial systems and IoT services.
Filip Popov (01:09)
Got it. Excellent. Excellent. So, you are very well immersed when it comes to the question of digitalization and IoT equipment and connectivity. So, to that end, I wanted to ask you in your experience, what are the top three things a company must do to prepare for effective connectivity before deploying an AI solution, given the fact that that is a prerequisite for any AI application deployment?
Adriano Costa (01:47)
OK, well, I think that companies need to respond to some questions before starting any projects, right? First of them is what is the challenge to be solved with AI, right? There are many things that you can do with this technology today. So, first thing is to respond to that and the second point in this question is what is the required data in the data flow, right? Because you need to respond to what kind, if you will get the data from streaming or from historical data and what's the flow because you can get data from different data sources in the industry, right? So, in the end, you can have the same effect. So, this is an important thing to check. A second thing is about security, right? You need to check security data access availability for the data and also reliability requirements. Because at some points you can have, I don't remember the word, the right word, but you have the holes in the data, right? So, you can be based on a thing that does not provide all data that you really need.
Filip Popov (03:15)
Incomplete data.
Adriano Costa (03:16)
And the last thing is about communication protocols, because you have many protocols available in the industrial environment. Think that you have different controllers, different SCADA, different systems. And for sure, in the side of AI, you have control of protocols you can implement, right? So, this is the side of the provider. But in the case of the industry, you need to know what protocols you have available and also in the cloud.
Filip Popov (03:49)
Got it, got it. You've actually, let's try to unpack everything you've said so far. I believe you started with saying you need to know your why, why you're even doing AI.
Adriano Costa (03:59)
Exactly.
Filip Popov (04:00)
To know which data you need and where you're getting it from. Make sure that it is complete for what you're trying to achieve. Cyber security.
And am I missing something else? And protocols, of course. So yeah, different protocols. And I suppose as you, as someone working in industry, you are a big fan of standardization when it comes to protocols across the entire shop floor. That makes sense. So, let's build off of that. We'll get to cyber security a little bit later. I also want to talk about that, but before we get into it, could you discuss one or two of the most common challenges companies face when implementing connectivity solutions and how they typically overcome them or how do you help them overcome them?
Adriano Costa (04:47)
Yeah, I think the first challenge in most industries that I have been attending in Brazil, Latin America, it's about the internet connectivity, right? So, thinking that IoT providers, I can separate in two categories, right? The solutions that can get the data directly from the existing systems, right? That get the data from PLCs, from SCADA, from different data sources. And also, the ones that are, let's think one more, one for database, right? Historical like SQL or Postgre or you can get this data and the ones that implement their own sensors in the field, right? I have been to many companies doing that because do it difficult to connect through the existing to the tools. So, I think this is the first challenge. No internet available in the industrial network in order to connect with IoT services. So, this is due to, I think, the security, data security, right? So do that. It's not something that a provider can change in the first moment. You need to, I'm going to say, understand during the, in my thinking, it's that you need to understand during the why and the data flow. That's why the first suggestion.
Filip Popov (06:23)
Excellent. Okay, fair enough. So, I'll kind of build off of that and we can segue into the conversation about cybersecurity. So how do you convince companies to embrace cloud solutions despite the security concerns? And can you share a success story and a specific story for the client?
Adriano Costa (06:37)
Yeah. Yeah, usually I think that it's difficult to convince customers, right? In my opinion, it's because actually in my experience along these years, you need to talk with different people in the industry to get the solution implemented, right? Because decisions are coming from a process engineer, from a production manager, from finance, from different areas. So, it's difficult to convince. So, I think that what I have been doing, I check the requirements. I build a solution trying to check the details of the premise, as I said in the first questions, and try to show the pros and cons of different architectures to show the customer that he can choose different solutions. After that, I try to solve, this is not possible always, right? So, I try to offer a proof of concept with a solution in order for customers to see the value. So, in order with the proof of concept, the customer who we are going to say the sponsor off our talk with this industry, we try to show the value of this proof of concept in order for other people in this decision can, OK, we can see this is a solution that works or this is attained our requirements in security because usually people that not see the things working go: "what is this?" This is not attended. This is against our policy. So, if people see the things working, I think maybe this is not easy for all the companies to implement, do a distance or things. So, I think in some cases we establish some templates, templates to solve basic things. So, after that, OK, I have experience with a steel company or, I don't know, a chemical company. You know the kind of issues or problems these companies have. And after that, they, how do I say, we can offer a template for this proof of concept. Right?
Filip Popov (08:58)
I see what you mean. So basically, productizing a type of service or product that you already delivered, you have success and you say, hey, I have this product/service that I can do. It's proven to work. Let me prove it to you and then we can scale it up. Is that how you typically overcome it?
Adriano Costa (09:14)
Yeah, exactly. Exactly. Because of the concerns of security, the concerns of the solution, you can, over many of these concerns, show the solution working and how, I'm not saying that it's easy, right? So, I'm saying that's a way to do it. Regarding the cases that you mentioned, I think was the second question, right? We have different cases with software based in the cloud, with IoT. But I think the solution, I can say, that can reach different developers of AI, I think, would be the use of OPC. The name that it uses is wrapper, right? Because let's think that usually, I'm talking about streaming, right? When you get the data in real time from the shop floor, right? So, in this way, let's think that you have different protocols. Usually people, many companies, they use OPCDA. It's based in DECON, but it's not so friendly to connect with different solutions in different levels, so that's why OPC UA is evolution. So, and there is a component that many companies have that's wrapper. In the case, we are a partner of a Matricon that has a Matricon OPC UA Tunneling. So, the feature of Convert OPC DA to UA is a way to access data with security, right? Because let's think that any profession that know Python, let's think this way, Python or a tool that uses Python can implement a library that transforms the solution to OPC UA client, right? So in this way, any AI solution that uses Python can access the data for OPC UA in the shop floor, right? So, this is the easy way to, we have helped many companies.
Filip Popov (11:20)
I understand. And in your experience, why do companies want this in the first place? Why do they want to live stream an immediate data accessibility? What is it? Why do they start these projects?
Adriano Costa (11:40)
Yeah. So I'm not a data synthesis, right? I studied some BI and things about, how can I say, data analytics, right? But in my understanding is that data synthesis needs immediate access for us to happily improve the AI models, right? In the side of who developed this, monitor the performance and the bug issues. I think this is the main motivation from the side of who developed AI, right? And I think this, how do I say, the end customer needs to provide data access to these professionals in order to, how can I say, first, how can I, how can I say this part? Let's think that companies need to ensure the access through road-based permissions., okay? It's a way to provide the access to these data sensors like we've secured. Also, they can implement, I'm going to say, edge data storage in some point of the company, in order for AI solutions to get data with security also, right? So, I don't know if I explained it right, if I respond to your question in the right way.
Filip Popov (13:17)
To an extent, I suspect. If I could summarize what you said, basically they need live stream data in order to affect the production in live, right? It's very important that they have the right information at the right time. And therefore, this kind of zero latency, if you will, is very important as to what the data scientists do that's kind of a little bit outside of your specialty. But from our experience, and what you've said is to increase efficiency of the production. But I think the name of the game is really cost reduction. You want to increase efficiency so that you use less stuff to make more stuff faster, cheaper.
Adriano Costa (13:58)
Yeah, exactly. That's it. Thank you. Thank you for translating.
Filip Popov (14:03)
That's how I read into it. In terms of what kind of clients do you see use these types of advanced analytics and data science in your, if you, if I don't know if you follow up with them after you've delivered your services of connectivity and integration, but I'm suspecting that's a lot of their motivation comes to connect in fact and transition to cloud is so that they can do data science and perform data science and perform advanced analytics. What types of companies do you find any commonality between them as to which companies do that in terms of size, industry or what they produce?
Adriano Costa (14:48)
Okay, we have been talking with different kinds of companies that are implementing. What changes between them is the, as I said in some moment, is the availability of connect with the web, right? So, some of the companies, mainly in the mid-size, the mid-size companies use like a database, right? So, they usually connect a database with data in the shop floor. And the solutions, how can I say, the services or IoT solutions can gather data from this database in the middle. Usually, the database is in the IT network, right? So that's a way. So this is the mid-site companies in different fields. Let's think that food, beverage, and I'm going say there is auto parts that, different companies use that. I'm going to say the big companies also companies that we work with, some many of them are using IoT services that use their own sensors, right? The providers of IoT solutions provide their own sensors so but they also connect with the systems when possible. So, because in these big companies also there are concerns about security. But sometimes they establish a computer in DMZ, like DMZ level, in order to be the connector from the shop floor to these IoT solutions. So, in the big companies, that's a practice, like chemical, paper, steel companies.
Filip Popov (16:48)
Understood. Basically, you're saying that this type of technology is not reserved for any type of industry and any kind of company, it will transition to cloud. You see in transitioning whether they're mid-sized or large companies, they just typically have different routes of how they get there and different suppliers.
Adriano Costa (17:06)
Yes, yes, exactly. So, what changes the architecture, right? Because as I mentioned at some point that, if you understand the why, the data type that you need and the data flow, you can establish a different architecture, right? So, I think any kind of company could use this AI cloud-based solutions.
Filip Popov (17:11)
Excellent, excellent. Okay, well we can make a little bit of an interlude. I brought my coffee because this is called Espresso 4 .0. So, I don't know if you have yours. Do you drink coffee? I mean, you're Brazilian, you have to drink coffee.
Adriano Costa (17:39)
Yes, yes, but I'll drink some later.
Filip Popov (17:45)
Sure, no problem. It is very early where you're calling from. So, we appreciate that you've actually tuned in without any caffeine. Congratulations on that. In terms of your business, who are your ideal clients? How would you describe them? What metrics do you use?
Adriano Costa (18:11)
Um, uh, could you, I have to tell you, I didn't understand the question. Sorry.
Filip Popov (18:16)
Yeah, who are your ideal clients in terms of size, readiness? Where are they on this journey? What are they looking over to solving? How do you help them? And lastly, how do you help them solve that?
Adriano Costa (18:23)
Okay, okay, okay, thank you. Mainly, let's think that we have been attending, in Inovax, we have been attending process industries, right? Usually, continue process industries. We have customers in the batch process, like food, beverage, but main companies come from the process industries. Yeah, I think it's the most of our customers. And yeah, oil and gas, paper, steel, yeah, process industry is our main target.
Filip Popov (19:07)
Process industry because they have a constant flow of data and therefore, they can also benefit a lot from having a good overview with and understanding it well, I'm suspecting. And last, but not least, before we part ways for this episode of Espresso 4 .0, what advice would you give to a company that's about to transition to cloud-based solutions and start doing AI or machine learning?
Adriano Costa (19:37)
I think define clear goals and expectations about AI. I think this is the understanding of what are the challenges they need to solve with AI, or problems that they need to solve. Because you can, I'm going to say, define AI solutions and set in some moment for different applications to solve different problems. And also, you have different architectural options. So, if you understand the why, maybe you can stand or approve, I'm gonna say, some, I'm gonna say some, some advance, like, a segmentation to connect the production to the web, or to connect authorization to change some security. So that's why I think I have a very clear statement of how your project can help to approve the initiative of AI Cloud Solution. And another thing specifically for providers of AI Cloud Solution would be implement OPC UI client, UA client in Dart Solutions, because I think we will make it easier in order to, I'm going to say, the the industries accept to connect because today OPC UA is inside the PLCs, inside SCADA, is inside the EMEA system. So, you can connect in different level if your expectation is streaming data, right? So that's my suggestions to the companies.
Filip Popov (21:24)
Thank you, Adriano. I intended that to be the last question, but now as you've answered it, I've opened up two more in my head. So, you've mentioned that different stakeholders need to be aligned in the company, particularly when it comes to the why, as well as the cybersecurity, as well as the fact that data can be siloed in different systems, right? Different departments. Now, in your experience, who drives this alignment? Is it a person or an entire department within the company, within the client? Or is that something that you do when you start doing a project? Do you align different stakeholders? Or do you work with one person, say, head of digitalization, who is the one that's thinking of the strategy, combining all of the information, getting all of the accesses and approvals to enable you to do the connection? So how does that work? Is that internal or is that something you provide?
Adriano Costa (22:32)
So, if I understand the question, in our case, usually we have a sponsor that already is leading a project of innovation, analytics or something like that. And he has the challenge, he or she has the challenge to connect with systems in the shop floor, right? So, usually, this is the main point that we have in our contact is the lead of the analytics project. That's if I understand.
Filip Popov (23:04)
Yeah, you did. You did. That's exactly what I was asking.
Adriano Costa (23:08)
Yeah, so yeah, it's the lead because we usually this sponsor is responsible to, I'm going to say, introduce the solutions. That's why usually people need a POC, a proof of concept. Like recently, we, just to mention a case, we have a connection with Google Cloud. So, we implement the solution in APOC, connecting with OPC UA servers. But in this specific case, the customer has a restriction because he needs to have a gateway in the industrial automation network and another one in the IT network. So, our two gateways provide the data to the Google Cloud. So, all the conversation was with a sponsor, right? So, and he, this sponsor, introduced our solution to, I'm gonna say, the entire team of innovation in this case that leads the, leads the, I'm gonna say, the innovation in different units of this company. So, now we have potential to implement the same solution in different plants. Do this approach, right?
Filip Popov (24:28)
Yeah, yeah. Okay, fair enough. So, there is someone internally that's driving that and that is typically the person that you're having the contact with, okay, that you're collaborating with. And the second question that came to mind is that you have mentioned the different types of architecture and even more importantly, different types of solutions that maybe mid-sized companies would prefer or more likely go with versus the larger companies that go with more established names. So, I wanted to understand in your experience, what would be the benefits for a company to build their own tech stack using different providers of technology, different smaller providers for like Risata, Inovex and so on, as opposed to going with, I don't know, a huge name, a giant like Siemens or Rockwell and so on. What would be some of the benefits of them tailoring their own thing from what's available as opposed to like, okay, I'll just go with these guys because I know them.
Adriano Costa (25:27)
Yeah, I think first is because these big companies are looking for big opportunities, right? So big projects. And many times, the customer needs to start small to understand the value of the solution, like I said, in some point with a proof of concept, right? And I think that companies that are, as you mentioned, can start with customers some small solution or can attain the specific needs and grow with him, right? Grow with this company and the solution, right? Can deliver a solution that can, I'm gonna say, deliver different benefits. So, and also it can be fast to deliver because a big company has, how can I say, different departments to approve just a change or something, you need to talk with many, many people inside the company to have something as the customer wants, right? So, I think that's the benefit of companies like yours, like Wizata, that can deliver this kind of AI solutions.
Filip Popov (26:56)
Got it, got it. So basically, flexibility and adaptability will be some of the main drivers. And I suspect also more attractive pricing often. OK, perfect. Well, Adrian, thank you very much for your time. Thank you very much for sharing your experience. We love to have you at Espresso 4.0. Maybe we'll meet up for another coffee sometime in the future. And yeah, if you have any closing words, now will be the time.
Adriano Costa (27:26)
Yeah, actually I'd like to thank you for the invitation again. It's great to be here and also talk with your audience. I hope that people can find some value in the experience that I shared. And if someone needs any information of our solution, just be in contact with us at inovexdigital.com.br – Distribuidora especializada dos softwares Matrikon, atvise e vNode no Brasil or it's easy to find me in LinkedIn too, right? So, thank you very much.
Filip Popov (28:00)
That's right. Absolutely. Have a great day. Bye.
Adriano Costa (28:04)
Okay, you too, bye bye.