In this episode, we spoke with Sundeep Ahluwalia, Chief Product Officer at TDK SensEI, about how AI is transforming data-driven decision-making in industrial environments. Sundeep shared his journey from data analytics to founding SensEI, and how his team helps manufacturers unlock the hidden potential in their operational data. We explored the challenges of AI adoption in traditional industries, the importance of domain knowledge, and how to move from descriptive analytics to actionable intelligence.
Key Insights:
• Plug-and-play AI: Automatic data labeling and model deployment with no manual setup.
• Edge-first design: AI runs on-device for real-time alerts without cloud latency.
• Predictive insights: Identifies failures weeks in advance to cut downtime.
• Enterprise-ready: Built-in cybersecurity, on-prem/cloud flexibility, ERP integration.
• Scalable platform: Designed to grow with use cases, from sensors to digital twins.
IoT ONE database: https://www.iotone.com/case-studies
Industrial IoT Spotlight podcast is produced by Asia Growth Partners (AGP): https://asiagrowthpartners.com/
Q&A Summary.
What inspired the development of the edgeRX platform, and what sets it apart from other industrial machine health monitoring systems?
The edgeRX platform was born from a clear vision: to deliver a scalable, low-touch AI solution for industrial machine health monitoring. Unlike many condition-monitoring tools that rely heavily on cloud processing or manual data labeling, edgeRX operates with full autonomy at the edge. The system includes vibration and temperature sensors, a Bluetooth-enabled gateway, and an autonomous AI engine. These components work together to collect data, generate and train models, and deploy inference directly on the sensor—without human intervention.
This edge-native design is crucial. Instead of streaming vast amounts of raw data to the cloud, which is costly and inefficient, edgeRX sends only the inference results. These are actionable alerts, such as anomaly scores, that offer real-time insight with minimal bandwidth requirements. The platform’s autonomous AI also self-trains using the sensor’s collected data, making it easier to scale across large deployments. So one thing that I want to highlight is that we have a low-touch edge AI experience for customers... install the sensor, connect it to the gateway and make sure the data is flowing. From there on, where our platform takes over, it collects the data, automatically labels the data, builds up the AI model, download it to the sensor, and off you go to the races. Combined with a user-friendly dashboard, edgeRX provides operational transparency and actionable intelligence for factory teams.
What are the core advantages of running AI directly at the edge compared to more traditional cloud-based analytics models?
There are two major advantages: speed and efficiency. First, by processing data at the edge, edgeRX enables real-time detection of anomalies. If a machine starts behaving abnormally, alerts can be instantly issued via email, SMS, or even directly into the customer’s existing dashboards or PLC systems. This helps maintenance teams act immediately or schedule downtime when it's most convenient.
Second, edge processing drastically reduces data transfer overhead. Instead of transmitting constant streams of vibration or temperature data, the sensor sends a simple inference result, such as a probability score indicating the likelihood of a failure. This makes the system more responsive and more secure, as sensitive operational data doesn’t need to leave the factory floor. In environments where network reliability or data privacy is a concern, this decentralized intelligence becomes a game-changer.
Do you have a real-world use case that demonstrates how edgeRX supports predictive maintenance?
Absolutely. In one project in China, edgeRX sensors were deployed across approximately 20 to 30 industrial pumps. Over a six-month period, vibration data revealed a slow, steady increase—an early indicator of mechanical degradation. The AI identified this trend weeks in advance and predicted a likely failure nearly two months before it occurred.
This allowed plant operators to plan maintenance well ahead of time. Not only did this prevent unscheduled downtime, but it also opened the door to deeper integration. Imagine connecting edgeRX with an ERP system—once the AI flags a likely failure, it could automatically check inventory for replacement parts and schedule the repair within a production cycle. That level of automation isn’t futuristic—it’s what edgeRX is designed to enable.
How does the platform utilize AI algorithms to build these predictive insights, and what’s the long-term vision for AI in this context?
The AI architecture is layered. On the sensor level, it performs localized inferencing based on vibration and temperature readings. On the gateway or edge node, it can aggregate additional inputs, such as machine telemetry, historical records, or trend charts. This hybrid edge intelligence enables a more deterministic approach to failure prediction.
Looking ahead, the ambition is to embed AI agents into the platform—autonomous systems capable of executing highly specific tasks, like monitoring a single production line or optimizing a specific type of machinery. These agents could use multimodal inputs beyond sensor data, including MES or PLC signals, to refine their outputs. Over time, edgeRX aims to evolve from a condition-monitoring tool into a full-fledged co-pilot for factory operations, capable of offering prescriptive—not just predictive—maintenance insights.
What are some of the standout features that make edgeRX particularly user-friendly and scalable?
The defining feature is its low-touch deployment. Customers aren’t expected to label data or develop their own AI models. Once a sensor is installed and connected to the gateway, the platform takes over: it collects data, labels it, builds the model, pushes it to the device, and starts inferencing.
That simplicity makes edgeRX highly scalable. Sensors are battery-powered and easily mountable—either via screws or proxies—making them suitable for diverse industrial environments. The platform also supports integration with SCADA, MES, and PLC systems, making it compatible with the broader ecosystem of industrial automation tools. One forward-looking feature being explored is a generative AI interface for the dashboard, where users could “chat” with the system to extract insights—turning analytics into a conversation rather than a spreadsheet.
What does a typical deployment process look like—from installation to full operation—and how is security handled?
Deployment is designed to be plug-and-play. The sensors are wireless and powered by long-life batteries, eliminating the need for complicated wiring or external power sources. After mounting the sensor, it pairs with the gateway via a simple switch mechanism. From there, the system begins collecting data, training models, and running live inferencing.
Security is embedded at the hardware level. Sensors are white-labeled and authenticated so that only verified devices can communicate with the gateway. This prevents spoofing or unauthorized data injection. Furthermore, customers can choose between cloud and on-prem deployments based on their data governance preferences.
How does edgeRX integrate with customers’ existing systems, such as ERP, MES, or dashboards?
Integration is key. Many customers already have dashboards and want alerts to appear there—not in an isolated platform. edgeRX accommodates this by offering flexible alert protocols. Users can configure triggers based on inference probabilities (e.g., send an alert if confidence is above 80%) and adjust these as AI performance improves.
Beyond alerts, the system can pull data from existing MES or SCADA systems to enrich its own models. This bidirectional flow supports both interoperability and continuous learning. Over time, as AI confidence grows and customer trust builds, deeper integrations—like automated ERP workflows or closed-loop maintenance—can be unlocked.
Where do you see this technology heading in the next few years? What’s next for edgeRX and AI in industrial environments?
We’re moving toward increasingly intelligent and autonomous systems. Today, edgeRX enables predictive maintenance. Tomorrow, it could enable prescriptive action: identifying the issue, suggesting the solution, and even coordinating the response. The rise of AI agents—tailored digital co-workers—will play a big role in this evolution.
Generative AI will also transform user interaction. Rather than clicking through dashboards, operators could use natural language to query machine states, ask for diagnostics, or even initiate actions. AI won’t replace human expertise, but it will augment it—accelerating decision-making, reducing errors, and maximizing uptime. The future of industrial AI isn’t just smarter machines—it’s smarter collaboration between humans and machines.
How do you see AI, IoT, and big data shaping the next chapter of industrial transformation?
We’re standing at the crossroads between Industry 4.0 and Industry 5.0, and AI, IoT, and big data are at the heart of that shift. These aren’t just buzzwords anymore—they are becoming foundational capabilities for future-ready manufacturing. Companies are now hiring AI leaders to orchestrate coherent strategies across business units, signaling how central AI has become to their transformation agendas.
At TDK, we designed our platform to support this evolution. Looking into the future, bringing in capabilities such as digital twin capabilities, AR/VR... I think the capabilities are endless. I can really see us leapfrogging into that area of providing all these capabilities on a single platform. Imagine a worker in front of a faulty machine being instantly provided with the schematics and guidance needed to fix it. That’s where we see the future: real-time, AI-assisted decision-making happening on the ground, at the edge.
With predictive maintenance becoming a fast-growing market, what is TDK SensAI's unique value proposition in this space?
Predictive maintenance is definitely a booming sector, but our approach sets us apart. While many players in condition-based monitoring started with basic sensor streaming and added intelligence later, we took the opposite route—we built our solution from the ground up with AI at its core. That decision gave us an early-mover advantage in model development and system design.
Another differentiator is our internal feedback loop. TDK runs many of its own factories, which gives us direct insights into what customers need. We know that early detection of machine anomalies—weeks in advance, not hours—is critical. This direct exposure to factory challenges helped us prioritize actionable insights and automation over generic dashboards. And because we started with AI, we’ve been able to design a scalable platform that can support not just hundreds but thousands of sensors and use cases.
How does the edgeRX platform fit into broader smart manufacturing trends and the transition to Industry 5.0?
The edgeRX platform was architected with scalability and modularity at its core. Beyond condition monitoring, it enables a spectrum of capabilities—from predictive and prescriptive analytics to ERP integrations. The vision isn’t just to monitor individual machines but to provide a comprehensive view of “factory health.”
This transition to Industry 5.0 brings new expectations: more collaboration between humans and machines, contextual intelligence delivered at the right time, and the ability to tailor insights to roles across the organization. edgeRX supports that by allowing companies to layer AI agents, digital twin services, and even maintenance marketplaces. Our aim is to make factories not just connected, but intelligent and adaptive to changes on the fly.
Many companies still struggle to scale AI in industrial settings. What are the core implementation challenges you’ve observed, and how are you solving them?
The number one challenge is data—its availability, consistency, and quality. AI thrives on data, but collecting the right kind across sensors, machines, and systems is hard. Most organizations underestimate the work needed to normalize and align this data for model training. That’s why we invested early in building our own models and pipelines, allowing us to iterate and mature our algorithms long before many competitors entered the space.
Scalability is the second major hurdle. Deploying AI to 1,500 sensors in a factory can’t be done manually. We developed edgeRX AI to automate model creation, deployment, and optimization at scale. And because it runs at the edge, much of the processing happens locally—either at the sensor level or via gateways—reducing latency and network dependencies. This makes real-time intelligence possible without overwhelming cloud infrastructure.
Security remains a major concern for AI in industrial environments. How does TDK address cybersecurity across its platform and deployments?
Cybersecurity is central to our design philosophy. From the beginning, we’ve worked closely with heavy industries—especially steel and manufacturing sectors—where data protection is paramount. These environments typically operate dual-network setups: one for business operations and another for automation, often completely air-gapped. Understanding this infrastructure helped us design solutions that fit into stringent security postures.
We’re fully aligned with SOC 2 compliance standards in the U.S., covering everything from intrusion detection to how we handle and store data. Every device we deploy—gateways, sensors—undergoes penetration testing. We close unnecessary ports, encrypt data both at rest and in transit, and sit behind the customer’s firewall to ensure multilayered protection. Even the data we collect is anonymized inference data, like anomaly scores, not sensitive PII or machine-level IP.
We’ve even had to complete four-page security audits for customers just to be considered for deployment. That diligence has helped us bake in robust security across not only the hardware, but the dashboard, login systems, and integrations—supporting features like MFA and SSO by default.
How do you see the evolution of TDK’s platform, and what new capabilities are you planning to introduce?
Our roadmap extends well beyond condition monitoring. We are already pushing into production health—identifying patterns and bottlenecks across workflows—and from there into full factory health. The goal is to move from “what’s wrong” to “how to fix it” through prescriptive analytics. And we want to integrate all this seamlessly into customer environments.
We also want to become a central hub for customers—offering not just dashboards but a true service ecosystem. Companies today use multiple platforms to manage maintenance, ERP, quality, and planning. Our vision is to unify these functions so that users have a single source of truth, enriched by AI. That includes expanding digital twin features, adding third-party services, and refining our insights engine to deliver actionable recommendations across roles—from operators to plant managers. We want to make sure that we are the one platform to keep track of what’s going on: overall factory health.
Ultimately, we’re building a platform that evolves with the customer. As they grow their AI maturity or adopt new manufacturing philosophies, edgeRX can evolve with them—securely, scalably, and intelligently.
Transcript.
Peter: Welcome, Sundeep. Would you mind briefly introducing yourself before we dive into the questions?
Sundeep: Well, thank you, Peter, for having me on the podcast. I really appreciate that. Hi. I'm Sundeep Ahluwalia. I'm the Chief Product Officer with TDK SensEI. SensEI is a part of the TDK Group of Companies. TDK is a well-known Japanese company with 49 billion in revenue. We were formed very specifically to address this condition-based monitoring factory health market with innovative AI solutions. As a background of myself, I've been in heavy industries for over a decade. I've been in industrial IoT space as well. So I've done both product management and business development over the course of my career. As a background, I have a double degree: Masters in Electrical Engineering and Masters in Computer Science. So I have a very well-rounded experience that I bring to TDK SensEI.
Peter: Excellent. Thank you, Sundeep. So let's go through the questions by starting with an introduction to edgeRX. Can you provide an overview on the edgeRX platform—how it integrates AI and edge computing for machine health monitoring?
Sundeep: Absolutely. So edgeRX is an industrial IoT machine health platform. It has several hardware and software components when we talk about a platform. The key emphasis that I place on platform is that a platform is designed to be scalable. As we grow with new use cases and opportunities, the platform grows with us as well. So it's designed from ground up to be scalable. In terms of the different components of this edgeRX platform, you have the sensors. For example, our edgeRX sensor has vibration and temperature, which is the most common type of condition-based monitoring sensors in the market today.
The key element of our sensor is that we actually run AI models on the sensor at the edge. The sensors then communicate with the gateway over Bluetooth. The gateway is basically a communication bridge that connects over Wi-Fi, Ethernet, or LTE to the cloud and for further AI processing capabilities. The other key component of our platform is our edgeRX AI engine. It's a completely autonomous AI engine that requires no human in the middle. It collects the data from the sensors, builds the AI model automatically and downloads it, and then the sensor starts the inference. Of course, another big part of our platform is dashboard. How do you actually take advanced analytics, informational insights? It's through the dashboard. For most customers, that's how they will see a product. It's on the dashboard. And so we provide live event data, as well as analytics, and any actual intelligence that will help you, your factories, in a healthy state.
Peter: Cool. Yeah, indeed, dashboard is the key. This is the key. This is the visible part to the customer, right?
Sundeep: Yes, the most visible part of it.
Peter: Exactly. Exactly, yeah. To come to the benefits of real-time monitoring, how does edgeRX improve the overall operational efficiency and reduce maintenance costs?
Sundeep: One of the nice things about running AI models on the edge is that we can provide real-time alerts. When we detect an anomaly, we can send an alert. We can send an alert through different mechanisms—through email, SMS. But also, we can send an alert over any machine interface to customers' own dashboards. All of these customers have their machine interfaces and things like that, which allows maybe a maintenance manager or supervisor to go in and take a look at, "Hey, what is the problem with that particular machine? Does it need to be addressed right away? Is it critical, or can it wait?" So it's one way that we can provide that.
The advantage of that is that they can assess the problem right away. If they don't deal with it, they know that the problem is there and then they can address it at a time that's more convenient for them—maybe a down cycle, maybe even when the system is not running at night time, things like that. So they can kind of address the problem at their own time. Now, that's a real-time aspect of it. But then you can imagine that over the deployment window, you are collecting data. You have trend charts. The biggest value comes from predictive maintenance, which is that I can tell you one month, two months ahead of time that, "Hey, I'm seeing a trend where this machine is going to fail." This allows them a good leeway to maybe find the part that needs to be replaced or schedule a maintenance time when it's more convenient for them. Because it's not failing overnight; it's going to take time. But at least, they are aware of it. And when that happens, that's when you have the most operational efficiency, and you lower the maintenance cost.
Peter: Makes sense. So the whole thing is running on an edge computing device, right?
Sundeep: That's right. That's correct.
Peter: I can imagine that there are quite some advantages over cloud-based solutions, like real-time monitoring and no delays basically. But from your point of view, what specific advantages does edge computing offer over cloud-based solutions?
Sundeep: One of the key things is, by running AI models on the edge, you're only getting inference data. A lot of our competitors, they stream data somewhere and then they use the data to do analytics and things like that. So you're streaming a lot of data to do for the analytics. With an AI model in the edge, you do all of that on the edge and all you send is an inference result. "Hey, maybe it's 0.75 or 0.65," whatever it is, just telling you the probability of what's going on with the machine. Or if you detect an anomaly, you're just giving them a high probability that this is a true anomaly, for example. One advantage is that it's real time. Second advantage is that you're not streaming data to the cloud for further analytics. Because that's a lot of data to send up. That's one of the, I think one of the key advantages of having AI on the edge.
Peter: Yeah, it makes sense indeed. Do you have a case study you could share, some hands-on examples, highlighting the benefits?
Sundeep: Yeah, so I think I can share one or two examples. I can't share any customer names. But I think one of the examples is that, actually, within China, we actually installed a large number of our edgeRX sensors on pumps. Sometimes pumps had one, sometimes pumps had two or three sensors on it. But over a period of six months that we collected data, we actually could clearly see from the data a trend chart where we could see that the vibration is slowly increasing over weeks. You can easily predict. Because we could easily predict two months before the machine fail that that is going to fail. That is the lead into predictive maintenance. Then you can take the leap. For the leap, it is saying, "Hey, I know that this is going to fail in the next amount of time, but I can even tell you what is going to fail just because now, over time, I've learned what it is. I can tell you what it is." Then perhaps in the future, I can link into your ERP system, tell you, "Hey, this part is in your inventory, on a scheduled maintenance cycle to fix it." Right? So I think I'm going to deliver it into the future. But with this particular example, clearly, the data we got from the sensors, we could show the use case that I can give you predictive analytics so that you can plan your maintenance ahead of time.
Peter: That's really interesting. So these were, yeah, personal interests. So it's pumps, right? It's motors, electric motors.
Sundeep: These were all pumps. These were all pumps. I think we had deployed over maybe 20 to 30 pumps or something like that.
Peter: Okay. Industrial major large size.
Sundeep: This is an industrial factory, yes, manufacturing.
Peter: I think the integration with an ERP, me being an SAP guy, that's actually an interesting thing to look into preventive maintenance in the plant maintenance setup. Yeah, very interesting, actually. Okay. So how does edgeRX utilize AI algorithms to enhance those predictive maintenance insights? Then what's the role you see for AI playing in the future?
Sundeep: So there's a couple of ways you can think about AI, right? We can definitely talk about it in AI on the sensor, but you can also talk about AI on the edge. Right? You have sensory inputs. Definitely, sensory inputs can tell you a lot of information, like I just mentioned, as a use case. But now when you start to look at other inputs that you can bring into your edge processing, even beyond the sensor, like at an edge AI gateway, for example—where I can now pull in machine data, maybe even pull in historical data, where I know based on trend charts what has happened in the past—now I can be more deterministic. Because now not only do I have my sensor data, but I have all this machine data that I can leverage at near real time and assess and be more confident or more deterministic about what's going on, and provide that analytics to the factory folks so they can go act on that.
To me, the capabilities of AI are endless, really. The more data you give it, the more smarter it becomes. And you know, the trend right now is about generative AI, AI agent. You hear this AI agent, agentic AI. All these terms are floating around now in the last several years. What we're seeing a fruition of that come to play, which is that more and more companies are launching with agentic AI, things like that. That's one of the areas where we are heavily investing into as well and AI agents into our platform, where they can do more focused tasks. Factory can say, "Look. Can you look at my process," for example? The nice thing about the edgeRX platform is not about just the sensor. It's about all the other AI capabilities that I can bring to a customer. Sensor is one input, but there's other inputs that I can leverage to build a more holistic view and provide them more information that can really give them that insight into what's going on in the production line, in the factories, and things like that. So the AI capabilities are endless.
Peter: Yeah, indeed.
Sundeep: I don't know if that answered your question.
Peter: I've been working in the MES field in the past as well, and there was always the demand for predictive maintenance. It was pre-AI edge, and it was never really reliable. It required a lot of calculations, a lot of sensors. It was not really reliable as well. So I believe that the whole AI pipe we have now, the whole development in the area of AI, will bring tremendous features to that area. Yeah, I can see that.
Sundeep: Absolutely. I think things are moving so fast.
Peter: Yes, indeed. So looking at edgeRx platform, what are the key features you have? What are the key features you offer? Let's say some like smart detection alerts, continuous monitoring. But apart from that, what are the key features?
Sundeep: Sure. One thing that I want to highlight is that we have a low-touch edge AI experience for customers. And that's very important, right? Because customers just want a solution to work. They don't want to go in, label data. They don't want to go build models for themselves. One of the key advantages we have is that we have built that capability into our system, into our platform—where you go in, you install the sensor, connect it to the gateway and make sure the data is flowing. From there on, where our platform takes over, it collects the data, automatically labels the data, builds up the AI model, download it to the sensor, and off you go to the races. Right? Now you start the inference engine on that, and you start getting inference results. So all of this, we want to make it seamless for our customers.
The other thing about this is that, we want to be able to scale. This allows us to scale very fast. Because I can now just deploy sensors everywhere and still automate the whole process of building a model and deploying a model and start running. So we take that on to ourselves and make it easy for our customers to deploy our solution. The other thing is that we have capabilities to now integrate with machine PLC, SCADA, and media systems. I talked about it earlier. Once you have that data input as well, then you can get into more of the AI predictive and prescriptive maintenance as well. That's a feature on our side as well. And of course, we talked a little bit about running models on the edge. The other thing is, you know, one of the things that we're linking on board as well is: how do you interact with the dashboard? Right now, it's very much like, "Collect this. Collect this, collect that." How about a generative AI version of it where you can kind of chat with the dashboard, and it'll give you what you need to see? Right? So these are all functions and capabilities that are part of what is called the edgeRX platform.
Peter: Interesting. That actually brings me now to the question of deployment as well. So how does the deployment process usually look like? What's the time frame? Then eventually, how does it integrate with existing solutions like we touched about, touched on ERP earlier?
Sundeep: Now, that's a great question. Our sensors are easy to deploy. They are battery-powered. One of the challenges we're going to factor is that power is hard to find. Sometimes cables come in the way of machine operation. So we've designed our sensors to have battery life, long battery life. There's multiple ways to install them. There's a screw onto the machine, depending on the machine type, or you can use a proxy to connect it to the machine itself. We made it very easy for our platform to connect to our gateway. So there's a little switch. You pair both the gateway and the sensor, and then off to the races you go.
We've taken security into mind as well, by the way. You know, all of our sensors are white labeled. So no random VLE sensor can just connect to our gateway and start sending data. So we have done that white labeling as well to make sure that only our sensors can talk, or only approved sensors can talk to our gateway. So once that is done, then I think, like I said, our whole experience of no-touch, low-touch comes in play. We have both options. One is of course the cloud. Many customers are hesitant to send data to the cloud. You can have an on-prem version as well to keep everything localized. But the key thing is that once everything is installed, then our edgeRX AI kicks in. We'll collect the data, build the model, download the model, and then put it into the operational mode for AI inference, and start the condition-based monitoring right away.
Peter: Interesting. So you have your own network connectivity, Wi-Fi connectivity, between the sensors?
Sundeep: Yeah, that's a great question. Our gateway supports Ethernet. It supports Wi-Fi, and it supports LTE as well. In the cases where customers don't want you to be on their network, we could actually go through LTE as well to the cloud.
Peter: Wow. Oh, that's cool. Okay. That's very interesting. I had another discussion with Sony producing LTE connectivity actually. So that's a place right into it. Interesting. Then in terms of integration, is it possible to integrate the system with existing enterprise software?
Sundeep: Yes, yes. So I think one of the biggest integration points we find is that customers want us to interact with their systems, with their machines. Because they have already many dashboards to look at. So we get this request all the time. "Hey, can you send an alert to me over my PLC to my dashboard?" Absolutely. We can make those interface points and work with the customers on what signals they want and at what level do they want it. We can make it agile in terms of like, a customer can say, "Look, in front of is the result. There's above 80%. Then send me the signal." In the future, they can change it. As models get better and better, the KPIs improve, they can say, "Well, let's make it 90%. When you see something with 90% deterministic probability, then send me that signal." So there's all that back and forth with customers, where that happens. And so our system is very agile and can make those adjustments. But the key thing is, yes, we can integrate SCADA, MES, ELC systems to get data from them, which we can use, and ASN signals to them that they can leverage and use.
Peter: Very smart.
(break)
If you're still listening, then you must take technology seriously. So I'd like to introduce you to our case study database. Think of it as a roadmap that can help you understand which use cases and technologies are being deployed in the world today. We have catalogued more than 9,000 case studies and are adding 500 per month with detailed tagging by industry and function. Our goal is to help you make better investment decisions that are backed by data. Check it out at the link below. And please share your thoughts. We'd love to hear how we can improve it. Thank you, and back to the show.
(interview)
Peter: Looking into the future, what features can we expect to be added?
Sundeep: That's a great question. If you look at kind of where we are focused and what's happening, AI, IoT, big data—all of these are buzzwords that are floating on right now. But they're here. They're here, and they're here to stay—those three things are now driving this transition from Industry 4.0 to Industry 5.0. We are right at the cusp of that. Many of the large customers are now putting in AI czars because they realize that they want to have to get into AI. They don't want all the different companies and divisions and factories using desperate AI solutions, so they're also thinking ahead about that as well. And so we play well into that, which is that I can bring you a very holistic solution that serves everything you need from a factory health perspective.
So looking into the future, bringing in capabilities such as I can bring in digital twin—some of the factories already have it, some don't—we can bring in digital twin capabilities. Cobots is becoming a buzzword, which is that you have humans and robots working together. You have like this whole gigantic AI, generative AI capabilities. Then one of the things that I'm really interested in is this whole AR/VR, which is that somebody on the field who's going to go fix that machine, can I send him the right data to easily fix whatever he needs to fix—schematics, whatever it may be? So I think the capabilities are endless. I can really see us leapfrogging into that area of providing all these capabilities on a single platform to our customers.
Peter: Interesting. Okay. Well, it's going to be interesting to see what comes in the future, I believe.
Sundeep: Yeah.
Peter: The shift from Industry 4.0 to 5.0, yeah, I think it's going to be an interesting one as well. People are already talking about it here as well.
Sundeep: Exactly. People are talking about it. They're making a strategy around that already.
Peter: Yeah. When it came, actually, this was a very, very hot topic here in China—Industry 4.0. And it's still being pushed by the government as well: modernize your manufacturing. It's a big thing. Maybe we could shift a little bit towards the market-relevant topic, away from the technical part. So looking at the market impact, given the rapid growth of predictive maintenance in the market, how do you see the competitive landscape here for TDK SensEI? Also, what kind of opportunities do you see?
Sundeep: Yeah, we believe that the market is really — the predictive maintenance market is really big. One of the nice advantages for us is that TDK itself has a lot of factories. A lot of inputs and feedback we get really talk about predictive maintenance and the need for as much as of advanced notice as possible. For many of our competitors — there's a lot of people in the condition-based monitoring space. We're not unique to that, with sensors and things like that. However, what's unique to us is that we are grounds up, started with AI. While many people just said, "Hey, let me stream my data, maybe use an ISO standard to detect what's going on," we just made the leap from ground zero to say, "Look, we're going to be an AI company, and we'll bring in AI from day one." I think we are ahead of the game in that respect. It's that we have started all our whole business based on AI. But the market opportunity is huge, and I really do think that leveraging AI will be an advantage for us in our growth.
Peter: Interesting. How do you see edgeRX align with the broader goals of — since we talked about Industry 4.0 before, how do you see this aligning with Industry 4.0? How do you envision the future for it and evolving into 5.0 and more intelligent manufacturing environment?
Sundeep: Absolutely. I think we can do condition-based monitoring with our sensors as well, right? But the leverage of AI really helps us differentiate that. The key thing, the key aspect in edgeRX platform is scalability. I can add on other services, add on other, bring in AI agents and things like that. So the platform has been designed to scale with not just a number of customers but by the number of use cases as well. And to really leverage the marketplace, I think we touched upon some of the things that I talked about before, twin, but other elements as well. Right? I can start looking at things like maintenance service, start looking at things like integrating with ERPs. These are all the capabilities that the platform has been designed—to bring that ease of integration and provide that one, single holistic solution to customers to focus on their factory health.
Peter: Okay, okay. I got a few more questions here at my end also, what challenges and solutions and the user experience. We have already touched on some of those topics.
Sundeep: We touched upon that, yeah.
Peter: If there is anything else you would like to add to what the challenges companies face today when they try implementing AI in their maintenance solutions?
Sundeep: Yeah, I think that the biggest challenge with AI is always data. The more data you get, the better the AI models become. So for everybody, the challenge is always getting that amount of data and different sources of data. So I think that's a unique challenge. It's not a unique challenge. It's a challenge everybody will face whoever wants to enter this market—any market that needs AI. That's one of the reasons we invested early so that we can start building our models and maturing our models well ahead of anybody else. The bigger challenge is really about scalability. I think we touched upon that earlier, which is that if you want to deploy 1,500 sensors, how do you scale that so that you can provide value right away to the customers? Right? That's when it's possible. That only becomes possible with a solution like we have developed, which is that we have this edgeRX AI that automates the whole AI model building and creation and deploying on our sensors at the edge. But we also have capabilities to do more processing at the edge, at the gateway level as well. I think some of those challenges are, I think everybody will face. But having been in this industry for a long time, heavy industries, these challenges we became aware of very early in my career, so we have tried to address that up front.
Peter: Yeah, very, very good. Okay. Because we have already answered most of those questions which I still had on my mind, maybe we can skip ahead to actually — you just mentioned cybersecurity considerations.
Sundeep: Yes.
Peter: This is always a big topic. I was once in the situation of having an interview with an AI, and I was reading the fine prints. Nah, I'm not going to go for it. I'm not going to go for it because that's just too much information going into the cloud, and I don't know where it ends up. This is for me as an end user standpoint. Of course, I'm also in a business. It's always a big question. It started years back when companies were worried about putting information into the cloud. Some companies still are. I'm trying to convince companies that, actually, it's safer up there than it's on your premise. Okay. So what is your take on this whole topic?
Sundeep: This topic is actually very close to my heart just because I've been dealing with several heavy industries—steel in particular. I did a lot of business in steel industry and in heavy manufacturing as well. There's one thing that's very common among all of them, regardless of the geography, which is that they all have cybersecurity concerns. IT is very good about keeping it pretty tight in their facilities on what is allowed, what is not allowed. Right? What we learned very early on is that, they have two networks mostly in the factories. One is the business network. Maybe they call it something else, but there's a business network that is of course connected to the internet. Then there's the automation or the engineering network that is an intranet and is not connected to the cloud or to any internet at all. That's one of the ways they're protected. We are very aware that, well, the last thing I want to be is the source of any cybersecurity attack. Right? So we take that very, very seriously.
So the few things we do, one of the things—I don't know if you're familiar with—the SOC2. In the US market, there's security and compliance controls that address a lot of these issues. We are already into that SOC2 compliance. That addresses many things, not just about the deployments of our hardware but it's also about the company—how do you handle any cybersecurity intrusion into our own systems, in our own company, as well as into customer environments? One of the things that includes is called penetration tests. One of the things we go through is that anything that we deploy like a gateway and things like that, we go and put it through a penetration test to make sure that all our ports are closed and not easily accessible to anybody outside. If you're on Wi-Fi or Internet, we are already sitting behind a customer's firewall. So somebody has to get to clear a firewall first. If somebody cracks a firewall, they have bigger problems than us. So they're very good about that.
The other thing is that we make sure that any data — first of all, we don't collect any personal identification information. It's only machine data. And even in machine data, it's just inference data, like 0.5, 0.95. Even if somebody gets that, what are they going to do with it? But more importantly, the data is encrypted at rest and in transit. So sometimes you have integration with customers' machine information, and that is proprietary information. We want to make sure that's protected. So we definitely make sure that everything is encrypted in the edge and in transit. Then, of course, I think you mentioned earlier like, hey, the cloud is actually a more secured thing, right? AWS cloud is quite secure. So we use all of the integrations with AWS to make sure that security is taken care of. But we go through that whole analysis to make sure that everything that we do is done in a very secure way. In the past, believe me, I have answered four pages of spreadsheet just on security for customers. So I know what they want and what they look for. So we have taken that into account, to make sure that we are compliant with all of the cybersecurity needs and that we are never the root cause of any penetration from a cybersecurity perspective.
Peter: Built with security in mind, that's very important.
Sundeep: Built with security in mind, yeah. Including the dashboard as well. We have a secured login auto. We can do multi-factor authentication. If customers want SSO, we can definitely put that in there too. We want to make sure that we are compliant and comfortable with what customers want to do.
Peter: Perfect. Excellent. Then let's come to a final question about the future. It's always the final question. What's the outlook? Looking ahead, where do you see edgeRX evolving into, and what are your plans for the future?
Sundeep: I think, first and foremost, we want to grow beyond this condition-based monitoring. We want to get into production health with predictive maintenance and predictive analytics, and then get beyond that into factory health with prescriptive analytics, provide more advanced and actionable intelligence to our customers—not just using our data but data from them as well—and really looking at how else can we serve those customers better on some of the examples I gave earlier about digital twin, other tools that we can bring on board. Already they use multiple platforms. We want to make sure that we are the one platform, that in addition to what data we may send to them on their dashboard, that we will bring value to them. A single point for all the different type of services that will help them not just keep track of what's going on in the factory, from the machine and production line health, but overall factory health, and take that across the board to other factories as well.
So with that in mind, it's how we created our edgeRX platform. Once again, scalability and security were the two key things that we looked at and we built out this platform. We look forward to how this industry is going to evolve. AI is going to definitely leapfrog for everybody. We can already see, talking to customers, they already have ideas and thoughts about what they want and how they want to use it. So some of the things we're thinking about are actually not that far-fetched. Customers are thinking about that.
Peter: Great. Thank you so much, Sundeep Ahluwalia. It's a pleasure. It was really a pleasure. Very interesting talking with you.
Sundeep: Same here, Peter. Thank you for having me. It was a great, great discussion.
Peter: Thank you.