Digitizing the factories of tomorrow: Deltia co-founder Max Fischer on making manufacturing more efficient

Intro:

Today, we’re talking to Max Fischer, founder of Deltia, a startup that allows factory operators to track, analyze and optimize their processes. From studying physical chemistry and quantum dots at ETH in Zürich to founding Switzerland’s biggest hackathon, today, Max has founded his second company that offers digital solutions in the world of manufacturing. Join us as he tells us all about his journey and what he hopes the impact of his work with Deltia will be. Read on to find out more.

From studying quantum materials in a lab to launching an AI-driven product to support factories, Max Fischer tells us how Deltia is tackling one of the biggest challenges of our time in manufacturing.

Could you tell me how you've got to where you are as a co-founder of Deltia?

I'm a mechanical engineer by education and studied at ETH in Zürich. I focused on something different at the end of my studies. I studied physical chemistry, very basic materials science for new materials, especially for photonic computers. I then started a PhD in this area at ETH too.

While I was studying, I, together with a few other people from the EU, started HackZürich, the largest hackathon in Switzerland, which was quite an inspiration for me. I got to see how fast you can move and do very complex things with software. Before that, it had just been me standing in a lab doing basic research with potential applications for the next 15 to 20 years in the future, which is exciting, but this hackathon was exciting in a different way.

I quit my PhD and started a company out of university called Actyx. I started the company with another friend who was already working in Industry 4.0, the digitalization of the manufacturing side of things, and also in the research institute. This was 2016, Industry 4.0 was a huge hype in Germany, everybody was talking about it, but nobody knew what it actually meant for factories.

We only knew the theory. We started off building customer software solutions for factories, more specifically for metal processing companies, chemical companies, etc. because this is what we always thought was interesting. We started to productize things that came out of the projects and eventually built a developer platform for other people in the space, especially automation engineers who wanted to digitize things around the machine, and we did a lot of solutions ourselves.

You were working on a PhD before you started your first company, what was that in?

It was in physical chemistry. So basic material science research on quantum materials called quantum dots.

So, very small nanoparticles which have quantum effects because they’re so small. You can use them for different applications, including photonic computers, and biometric applications.

You went from the PhD straight into the company. Did you get frustrated with how slow academia can be and feel more excited by the pace of entrepreneurship?

I think academia is very important and it’s important that people spend time on that. But yeah, was a different type of thing. I was like, I think I was working with software, so then you can just like have a lot of impact very fast. I was more excited about that.

How did you go from working in science to working with manufacturing and production? What was the link there?

During my engineering education, I always had some touch points with the world of production and I had to learn the basics. I also had a friend from ETH who was in this digitalization of the manufacturing side of things, working in a research institute. We came together and started working in this area.

Can you tell me how you got into working with AI and production?

In my previous company, I digitized a lot of factories from different industries with non-AI types of solutions.

With a tablet, or with tools that you had to integrate into machines to understand whether the machine is running or not. Then you had business intelligence tools, which allow you to visualize the data and provide visibility. But with all the companies that we worked with, they all had a problem understanding what was happening on the shop floor.

So you have a lot of patience. Sometimes you have a good day, sometimes you have a bad day, but it's really hard to understand why it was good, or why it was bad, or how you can improve. This was especially the case when many things on the shop floor, were not digital yet. I saw that it was very tedious to get this information. It’s even difficult with things like tablet applications for workers because you have to train the person. Getting data manually is also unreliable because people, of course, forget to enter some information sometimes.

You don't have a lot of context about what is happening. Even if somebody says “Okay, set up a machine,” you might know how long it took, but you don't know what happened exactly, why it took longer, and what the problems were.

Also, even if you have the data, it’s very difficult and painful to understand how to make sense of it. How can you improve a lot of know-how that’s not digitized, especially the human side of things that are not digitized for factories?

And that's sort of triggered me to think about a new product. Can you not use AI to help with that?

I joined Merantix in Berlin, which is very well-known in the AI space, and a lot of people who know a lot about AI are there. So, I went there and we explored a few different ideas together.

My co-founder Silviu comes from the autonomous driving space, so he knows a lot about computer vision and artificial perception.

Computer vision looked like a great solution to digitize shop-floor processes; Silviu had some touchpoints with manufacturing so he was immediately excited about the space as well.

So, how does Delta then use AI to help people who run factories, how do they go step-by-step?

The problem that we had to effectively solve was overcoming the shortage of skilled labor. People who are not from manufacturing might have the idea that factories are super automated, that you just put in raw material, everything is automated and then the final product comes out, which is not the case. For most of the factories, you have an extremely important human component. People are involved in assembling the product, repairing the machine, and setting up the machine. But the problem is that these processors are usually the least digitized ones.

Right now, in Europe, we have the problem that a lot of people are retiring, so a lot of know-how is being lost when people leave the factory because this know-how isn’t being digitized. In the US a lot of manufacturing facilities are being built since the US is trying to become less dependent on China, but you don’t have a lot of experienced people who have worked in factories before

Now, we are addressing the problem in three phases: the first one, and this is what you today is basically, we improve the efficiency and the effectiveness of these processes.

We do that by digitizing the manual shopfloor processes with computer vision. This means we have cameras that are observing production lines, and production stations, and then we have an AI system that automatically recognizes processes i.e. somebody is assembling a product at the station or a product is flowing through a production line. We can show inefficiencies in the process and offer better ways of working. We are also documenting how the process is being conducted and what the best practices are.

We are digitizing the know-how of the most experienced workers. This can then be used at a second stage to train and assist the more inexperienced people in learning these processes and can offer them feedback about whether they are doing things correctly or efficiently.

This means that you don't have to have these very skilled people to execute specific processes like assembling a gearbox, for example. But you can use slightly less qualified or less experienced people because you can train them on the job and execute the processes more effectively.

Lastly, there's a mid-term goal for us because all of this will not be sufficient in the face of the size of the shortage of skilled labor in the next few years. We need to start automating this process, so we will also use our system that recognizes how a process is conducted and we are using this information to teach robots how to execute some of these processes.

Will robots replace human workers completely? What is the role of human employees in the factory of the future?

Employees will have to work with the technology for sure. But we also believe very strongly that we are making work safer and better for people too. There are a lot of people who don't want to work in factories anymore because it is a very stressful job. From a health perspective too, sometimes factory jobs aren’t very good. One-quarter of all assembly workers, for example, complain about heavy back pain, because it is physical work and strenuous for your body.

These are also things that we are observing. So we are not only looking at efficiency but also at the most sustainable way of working. Looking at the ergonomics, for example, can you improve the process in a way that has less strain on the worker? And can you help them to do these jobs longer so that you can keep people in the job for a longer time?

And then assist as soon as the worker is doing that. Of course, there, you will eventually automate more and more, but it's more of a co-existence. So I don't see that the factories will be fully automated for a very, very long time, it's more about the robots doing heavy tasks and more complex tasks being done by humans. You use the ability and the creativity of the people to do other more complex tasks, but it's more a co-existence for sure.

And then, AI will help the workers too.

Your co-founder has experience in autonomous cars. Is the technology, then, that you are using i.e. the cameras that are tracking processes in the factories, has that been borrowed from the autonomous car industry? Is it a new use for this technology?

It's hard to drive around the streets. For us, it's more like a static case. You have a camera, for example, at a station. But in the way that it needs to perceive the environment, also there are a lot of things that are changing constantly in a production environment, there are different products, for example. And every production line, even within the same factory, is tasked with making a different product. So the technology needs to be able to cope with these variances, Sometimes you have changes in the layout, so you need a very flexible system which, like a car that drives in the streets and you go from Germany to Austria and the environment is a bit different, you need to have a system that adapts, so yes there are similarities between the streets and the production environment.

What does the future look like for Deltia? What's your big vision for the next five years?

There's a huge opportunity in the manufacturing industry right now. There are a lot of micro trends which of course factories are adopting to digitize faster. There will be an even stronger focus on efficiency and automation.

Video analytics is a very novel approach. The basic technology is there, but it's not an established category yet. There's an opportunity for Deltia to become the leading company in video analytics and use computer vision as a way of digitizing short-flow processes and making short-flow processes better. That’s something that we will work on for the next few years.

There's a big push generally across the world to automate manufacturing, so how are you positioning yourselves as a company?

There are several factors contributing to this push right now. The demographic change is a huge drive in Europe because people are retiring and you have a lot of experienced people leaving the workplace so we need to do something about this. This is something that a lot of our customers are telling us. There is an existential threat, to a certain degree. If 20, 30, or 40% of your workforce is going to retire in the next five to eight years, that’s critical, right? And then you don't have a lot of new people coming who want to work in this space. It's more difficult than ever to find qualified personnel.

In terms of the macro trend, if you look at the US, for example, there are a lot of facilities and factories in general being built up because a lot of production is being moved back from Asia. Because of COVID, people have realized that it's a good idea to have a more decentralized supply chain and not to have everything being produced in China or India. But in the US, almost an entire generation of know-how has been lost as people start to move manufacturing away in the last 20-30 years.

Manufacturing was strong in the USA, but everything moved to Asia and now it's coming back. But right now, even though production is coming back, there aren’t the same amount of qualified people to work in the factories, so factory owners can use our technology to drive efficiency and improve the quality of their products.

Are you focusing mainly on Europe?

Today, Europe is of course strong, especially in Germany there are a lot of very successful manufacturing companies. But we are also selling to the US, we also have customers in the US. Since we are a German company it makes sense for us to start in Germany, but as we are approaching the growth phase it makes sense to sell to the US because so much is happening there right now.

What kind of customers do you have?

We have quite a few interesting customers. Most of our customers are from the machine building and electronics industry. So we have quite a lot of manufacturers of heat pumps, for example, construction tool manufacturers, power tools. They work in different industries but they’re mainly in machine building and electronics.

Will your product have to change for different types of factories or is it quite generic?

The way that we generate data is quite generic. In essence, what our system is doing is trying to identify a process that is a recurring pattern over time. It’s something that happens over and over again. So if you observe a station or a production line and you just look at it for a day or a few days, you will always see things happening similarly. A product comes in, a product goes out, a door is opened, a door is closed, a light goes on, and a light goes off… etc.

These are patterns that our system can automatically recognize and you just have to put a label and a name to these patterns, or processes. This can be done by the customer to generate this information. It's quite generic and can be used across industries and factories.

It’s not even just factories, we’re also looking into using our product in logistics and other industries. What is more interesting are the applications on top of the data which are more use-case specific. For example, today we are focusing a lot on manual assembly. It always works similarly. You have several stations and you have a product that is being worked on at each station.

If you want to understand how long people are working on a product at each station or what the lead time for the product is for each line, then there's always similar information you can analyze. It’s just more use-case-specific. If you look at a fully automated line, for example, then you will see something different.

How do your customers use your product? How do you marry the software and hardware components together?

The user in the end sees a web-based analytics application to analyze production data analytics application. The camera technology is in some sense hidden from the user since it works in the background - the AI continuously analyzes the video streams to create process statistics. At the end, the user only sees the process statistics. The cameras are used as a sensor, which is why we often frame it as a “Cameras-as-a-Sensor” approach.

Do your cameras track the different people working within the production?

Our cameras are looking at the process, not at the person carrying out the process. We don't identify people, we don't generate any biometrics information, so you can't know who operated it.

Sometimes our customers might be interested in videos for training purposes, or for documentation purposes. But if you do have a video snippet available, you can cut out the person completely and make sure that you can't recognize a person in the videos or the data. It's not about the individual, it's about the process. It enables our customers to ask themselves whether they need to change something in the process to allow their operators to work better. It's not about the performance of individual operators.

You founded another company before Deltia that was also involved in manufacturing, what did you learn from that experience that you’ve taken into your new venture?

My former company still exists and it also involved working on the digitization of the shopfloor, but more on the infrastructure side of things. It was more like a database and oriented towards developers and automation engineers who wanted to build solutions for factories to digitize things around the machine. And they could use our infrastructure to build these solutions.

I left the company in 2022. At its peak, we were around 35 people and we raised around $10 million in venture capital. I learned a lot about building a company. About hiring – when to hire and who to hire? How to build a team, how to motivate teams, but also I learned a lot about manufacturing. In manufacturing, it's really important to focus on very narrow use cases because there are so many opportunities you can follow in manufacturing.

I also learned how important ROI is. It's like incredibly important for manufacturing companies. They are a bit tight on money because it's a low-margin business in general. They are also willing to invest a lot of money, if there’s ROI. They buy machines for millions of euros, or build factories for hundreds of millions of euros, but always need to see a clear ROI case.

What does this launch mean for you now?

The first phase of our company is building our analytics application. We started very early on in the design process working with customers to build our product. It was very important for us to build something together with our customers and solve real problems for them.

We started building the product last year, so we spent that year building the first version of the product and validating that the technology works and validating that our customers get value out of the system.

And I think we’re now at the point where we have this version of the product that has been validated with quite a few customers. This now brings us from the first half dozen customers to the next 2 to 3 dozen customers in the next 18 months.

What do you think the biggest impact of your product is going to be in the future?

I think it will have a big impact on helping factory owners manage the shortage of skilled labor and provide more sustainable and healthy processes to workers. It will become much easier for less experienced workers to learn tasks, get support, and have the flexibility to learn on the job. Factories will no longer have to be afraid that when somebody leaves, that they won’t be able to operate anymore.

Sometimes they only have one person who knows how to operate one machine. And I think that's a dangerous situation for factories because they need to find a more sustainable way of running their operations.

Our product will also make factory and production work much more flexible so that you can learn your tasks more easily than in the past. I also think that people will be able to stay in their jobs longer.

Our product will also have a big impact on the health of employees by ensuring that people are not as stressed and they don’t have to carry out heavy manual tasks anymore.

Generally, people don't realize how important factories are. Everybody's talking about electric mobility climate change, wind turbines, solar modules, etc.

But everything comes from a factory. So everything that we need for the energy sector, for electric mobility, everything comes from a factory, and if we don't have people in factories these things can’t be produced. We can't automate everything, we need the people to be working in the factories, building the products. We want to solve a lot of the problems that we have in society. Everything that you touch comes from a factory, so we need to focus on this industry.

Are you fundraising right now?

We did a €4.5M seed round last year and will use the funds to build and extend the product, making it easier to scale and install the software in the factory, and covering more use cases. And then of course, we’ve focused on the commercial side of things and we’re bringing our solution to more factories.

Thank you very much for your time.

Thank you!