Enabling low-code development with AI Transcript – Part 2
![](https://blogs.sw.siemens.com/wp-content/uploads/sites/19/2024/07/Siemens-mx-mendix-02-002-_original.jpg)
Recently, Subba Rao, Director of Manufacturing Industries Cloud for Mendix, talked about the applications of AI in industry and, specifically the creation of AI-augmented low-code applications. Check out the full podcast here or read along with the transcript below.
Spencer Acain: Hello, and welcome to the AI Spectrum podcast. I’m your host, Spencer Acain. In this series, we talk to experts from all across Siemens about a wide range of AI topics and how they apply to different technologies. Today I am joined once again by Subba Rao, Director of Manufacturing Industries Cloud for Mendix. Picking up from where we left off in the previous episode, you’ve been kind of talking more about AI assistance, right?
Subba Rao: That is true, yes.
Spencer Acain: So what about the AI augmented applications? How do those differ from the AI assistance?
Subba Rao: Yeah, this is a little bit different. I mean, assistance is more for the developer and their productivity, but like in the industrial space, I’ll people to the industrial space, right? So AI and machine learning is nothing. It’s been there for quite some time for several years, I should say, in the industrial manufacturing space. There are a lot of custom machine learning algorithms or AI models are built for different use cases, whether it might be in the quality or it might be in the maintenance side towards like you see using vision systems and data from the vision. There are already like machine learning is been applied to predict the quality issues or even like looking at the sensors and IoT and other data that’s kind of like attached to the machines. There are models, machine learning models that are applied to get into the productivity of maintenance of the machines, right? So these are all, it’s already existing. And at the same time, like a lot of our ecosystem partners have like several cognitive services that industrial manufacturers have either leveraged or planning to leverage. What does it mean? And the partners like AWS, there are several like cognitive services, whether it might be Amazon recognition or text track or poly or even like a bedrock generative AI services, right? So all these are like services, AI related services that are available. And at the same time, the manufacturers who have been used to kind of experimenting and trying out like AI and machine learning models, they’re looking at like a how we can leverage those and make our applications more smarter, right? The applications when you build an applications, it is for certain purpose and for certain user to consume it, right? So how do you make those applications more smarter by leveraging all this either the AI and machine learning models that they already built or they already like tried within their certain use cases or the cognitive services that they they’re planning to leverage from our ecosystem partners or hyperscalers like AWS or Microsoft or Google, right? So with Mendix AI, augmented capabilities, right? Users can easily connect to any kind of third party services like to interact with AI or machine learning models. So what does it mean, right? So just as I said, like AI and machine learning is nothing new for industrial manufacturers, they have been doing, but it’s kind of like a smaller scale in some cases, in some cases, okay? It’s an advanced complex use cases, right? So if it slides that further, let me start with machine learning, right? So enterprises like within Mendix, there’s something called machine learning kit. So enterprises can use the vision like they’re like proprietary machine learning models that they have used in some of their use cases or they’re planning to use, with cognitive services. And then with the machine learning kit, they can easily embed the machine learning models that are like from these different services, right? And the results of those machine learning models, right, can be easily applied into the Mendix application. So what do I mean by that, right? So there’s models that are trained with different frameworks, like either it might be the services from AWS or the custom services like TensorFlow, CAF or PyTorch, right? There’s all like different machine learning models that typically they kind of like provide an output call like ONNX, it’s open-natural neural network exchange, mouthful. It’s an open standard for machine learning models. So that output can be like using machine learning, Mendix machine learning kit, you can embed those results into your application and like how visual layer, right? For the users, the more visual it is, the more they can act on it. And that’s what you provide with the combination of leveraging this, the results of this machine learning models and embed that into your applications, make the applications smarter all with simpler, visual development mechanics that Mendix provides. Earlier I talked about like a Mendix development environment is all like visual development. So even these machine learning models embedding that using this machine learning kit, it’s a simpler drag-and-draw, right? That’s the first aspect with the machine learning. And similarly, like there are like a lot of AI services, I mentioned like especially if I take an example of our ecosystem partners like AWS, right? There are like Amazon Textract, Amazon Recognition or like IoT Filmmaker or Poly or Translate all these like different services that are there from our ecosystem partners. So Mendex makes it easier for teams to integrate those services in again the simpler drag-and-draw. So there are like several connectors available for these services. So by having those connectors, it’s available in the Mendix development environment where the developers can simply drag these services and then use them into their application logic. So this kind of like provides additional flexibility and ease of use in embracing these services and at the same time using the data because anytime we talk about AI or machine learning without data, there is means you cannot apply these, right? So able to make the data accessible to consumers who are using these applications. So that is the combination that we are talking. And lastly, like I talked about all these cognitive services, generative AI, which we have been talking, we have been seeing that more predominantly in the industry nowadays, right? So even that one like with AWS Bedrock or OpenAI connectors, it makes it easier for industrial users to compose or build their applications at ease. So obviously in the industrial context, the applications vary, whether it might be a machine learning application or AI application or gen AI, right? So it’s all dictated by the use case. You cannot say that, okay, because there is gen AI, we are hearing about gen AI today, all the applications will fit into that. That’s not the case.
Spencer Acain: I see. So I think we’ve talked quite a bit now about how you’re doing this and what you’re achieving through Mendix both in assisting developers and the applications that they can develop with AI. But what are some of the like use cases and uses of these AI and ML and generative AI applications and capabilities within industrial context, such as manufacturing?
Subba Rao: Yeah, that’s a good point. So especially when it’s like there’s a lot of talk about AI, nowadays we are hearing the practical use of these like the capabilities that I talked, the manufacturing realm is when you apply it in the right way and empower the organization, that’s where you’ll start ripping the benefits, right? So in our experience and especially engaging with our customers, we see there are several digital initiatives that are going on and like many of those digital initiatives are to solve certain problem, to solve certain like challenges that our customers are facing. And just to slice that further, the way we see it is all those use cases or digital initiatives are falling under like three to four different categories. So what do I mean by that? One of them is more about optimization, optimization of the processes that might be on the manufacturing side or supply chain side or the cross domain, right? Or even prediction side you want to kind of like getting into the prediction. I talked briefly earlier about quality prediction or maintenance prediction of your assets or even like the demand predictions, right? And then third one could get into like discovery side of the things. So there is so much data over there to discovery of the data and knowledge exploration. And lastly the transparency of all these things across the organization and also to their supply chain, right? So all needs to be coordinated with data. And if you are talking about manufacturing, there are like a lot of devices on the manufacturing floor, people and processes. This is where like AI, machine learning and gen AI plays a key role. So if you look into the individual aspects of it and the use cases, right? Let me start with embedding the AI modules into the applications. So there are like in the manufacturing set up, there are a lot of material movement that must be orchestrated to drive optimization. What do I mean by that? By that. Within the shop floor in the manufacturing does, if it’s a flexible manufacturing set up, right? So there are a lot of materials that gets more the AGVs or like people moving the material, getting into the station at the right time. There is a dependency with the warehouse, pardon the material, the availability is, and then like depending upon like what you are producing, the needs for the material could be different, right? So it’s a complex endure. There’s a lot of like different systems and people interaction that goes on to make it like a smooth in the operations and with the power of simulations and the combination of the real time data that is coming from the shop floor across all these processes, right? So you can apply the AI models to try the optimization. So like traditionally, there is several organizations are already using simulation models to optimize, to perform the what if scenarios, right? But like if it is just like a similar synthetic data in the simulation models, just imagine the power of combining that with the real data that is coming from your machines, your sensors or the actions on the shop floor with the paper, right? Combine that with your simulation data, the synthetic data with the real data and apply your AI models, and provide the flexibility to the users. Obviously, there is an building an application that I talked earlier and then using that application. So not everybody is a simulation expert, not everybody needs to kind of like be expert in data science to analyze the data. By having an applications that is consumable by the user with simple parameters, simple data inputs, right? They can perform the what if scenarios behind the scenes, all the simulation is running with real time data and combination of synthetic data. And with those what if scenarios, results, right? They can users can take the right actions. And this is a perfect example on the production flow optimization use cases that we see where the combination of simulation data, the real time data and applying the AI models, right, is ripping the benefit. So that is one example of embedding AI models into the applications because the further users, they don’t see all the AI models behind the scenes, they don’t see all the complexity or all the like advancements that are done on the simulation side or capturing the data from the real time. It’s more giving the right inputs and reviewing the results and taking actions from it, right? That is that is one one aspect. And second one is machine learning applications. I talked earlier about like machine learning kits with that’s available in Mendix. So take an example of a predictive quality or predictive maintenance. So typically, it’s to get to the productivity, it’s highly dependent on analysis of relevant data and learning from it. So the different machine learning models are already as I mentioned already used and used by the manufacturers. So what we see is using this machine learning kit, those machine learning models are kind of like embedded into the application and the users will get the trends. So there is a machine that’s kind of like getting monitored and with the data that is coming in, yes, they get the trends and with the trends, applying this machine learning models, it can even get into the visually within the application, give indication. Okay, there is a trend of one of the parameters going a little bit like in a different way. So it needs to be monitored further and it leads link to the prediction in a more visual way within the application so that users can act on it, right? So that is the second example and the third one is generative AI, right? So when you talk about generative AI, it’s more the, as I said, there’s a lot of data that is required for any AI to work with generative AI, it’s towards like a knowledge exploration of this data. So there’s a lot of complexity in the data that’s and the correlation that is required to make a sense of the data, right? So that is a knowledge exploration part towards the conversational experiences. So what I mean by that is able to converse with your system in a simpler natural language, right? To get meaningful information at the same time, not just getting the textual information, that is the beauty of generative AI, able to generate new content from that as well. Either it might be a text content or it might be like especially in the industrial context, driving some generating a plot, right? A graph. So those are the things that we are seeing. So here if I slice that further, getting meaningful insights, from this complex data that is spread across different different domains within an organization is where generative AI really helps the domain experts who don’t have time to slice into all this data and understand different graphs that they’re used to in terms of reports that they receive in a much simpler way, right? So that is that is one key aspect of generative AI. Here one good example is one of our customer has like leveraged the generative AI and created a virtual engineer that helps their quality and production personnel, right? Who typically for the quality people resolving the complaints, resolving like any issues that are coming either from the field or even internally depending upon the complexity of the product, it takes time, right? So by using the power of generative AI and then application that is built with Mendix, obviously with all those generative AI services that are coming from in this case, it’s AWS bedrock, right? That’s making it easier for the quality department to resolve the complaints which traditionally it was taking like more than five days to less than a day, right? Just because they’re able to like use the natural language, query the data and get that insights. At the same time, the production issues, this is on the quality side, the production issues which might be either the process issues or it might be the machine-related issues or it might be training issues, all these are kind of like the data that is there. Just getting to that is always challenging. So with the natural language with this customer that I mentioned, they were able to reduce this, identifying the causes of the production issues from hours. It used to take hours, sometimes a whole shift to less than one hour, right? So that is the power, that is the value that you see in terms of like how generative AI is getting embraced, but for that obviously data is very key element. Interesting. So it sounds like you’ve got quite a few applications here and you’ve really kind of honed in on that bringing that use of use and pulling those sufficiently long times down, you know, like you said, could take five days down to one for some of these sort of applications.
Spencer Acain: And on that exciting note that this is all the time we have for this episode. Once again, I have been your host joined by Subba Rao on AI Spectrum podcast. Tune in again next time as I continue my discussion with Subba on the many applications of AI within Mendex and how it’s changing the landscape of industrial software.
Siemens Digital Industries Software helps organizations of all sizes digitally transform using software, hardware and services from the Siemens Xcelerator business platform. Siemens’ software and the comprehensive digital twin enable companies to optimize their design, engineering and manufacturing processes to turn today’s ideas into the sustainable products of the future. From chips to entire systems, from product to process, across all industries. Siemens Digital Industries Software – Accelerating transformation.