Case studies Cerrion

Cerrion: Empowering Production Teams with AI-Powered Anomaly Detection

Karim Saleh, CEO & Founder of Cerrion, shares how his startup is solving a problem in manufacturing through AI technology, working with LEXR, and his advice for the industrial sector.

Cerrion

I am Karim, the founder of Cerrion. I’m originally from Egypt, where I grew up, and first came to Zurich to study at ETH. Immediately after finishing my studies at ETH, I started talking to various manufacturing companies about the possibilities of achieving exponential growth in the industrial sector through technologies such as AI, computer vision, and more. We very quickly identified the fact that the industry still has a transparency problem, and that’s really where we started.

In my case, I asked people directly. About a year before Cerrion was founded, I spent my time talking to different manufacturers and the various stakeholders involved. I found out what their challenges were, what the status quo was, where technology could fill a gap, and how significant that pain point was. Together with the team, we then identified a problem we could solve, and initially ran consulting projects where I tested different approaches. Eventually, we found the approach that would allow us to solve the problem with technology. We also clearly saw a market pull, in the sense that customers were willing to pay us for projects even without us having a product, purely on the basis of the vision. That was a strong signal for us to go all in.

We quickly learned that things can go wrong very fast in production. Processes have become extremely complex over time, and with so many processes and machines involved, it’s easy for something to go wrong. The challenge is keeping track of where things are going wrong. As a result, you always need someone constantly there to observe what has actually happened and why. So we said that instead of having a person constantly watching everything - which no one can do given their limited attention span - we would try to replicate that task with AI. Humans observe processes with their eyes, which can also be done with standard video cameras. These cameras are then installed at various points throughout the production process. The video streams from the cameras are then analysed by the AI. The AI learns what the process looks like and can automatically detect anomalies and classify them into different types of issues based on simple human feedback. Depending on the type of problem, we can then decide whether to stop the machine or the affected section. This allows us to reduce safety risks, increase efficiency, and save the customer money. It also means a worker doesn’t always need to be on site to monitor the process, enabling a more efficient use of labour. We’ve also developed a platform where all the data is stored. This allows production teams to optimise production and decide what to focus on.

Manufacturing in general is very hectic - there are always 1’000 fires to put out. The difficulty for us wasn’t building a vision technology that could recognise problems, but getting that value to people in the simplest way possible. How do we build interfaces that make it very easy to capture that value? For example, we’ve built a ticketing system that tells people: look, here’s a problematic pattern, here are the videos and all the information you need, now do something about it. We also suggest what can be done to fix the problem, and then tell you whether the problem has actually been resolved. Our biggest challenge has been understanding how the industry works today and figuring out how we can fit in without being seen as just another system. We did this by spending a lot of time with our customers. We have customers in different countries, and we travelled to them to iterate and gather feedback. Technologically, AI wasn’t trivial either - the solution lies in having a good team capable of doing it.

We’re working on making Cerrion even faster and more reliable. There are several things that need to be addressed so that production lines can run unattended when Cerrion is in place. Another focus for us is making the AI more transparent, providing more governance around it, and developing the whole system further towards self-service. We already have a clear roadmap for all of this. We’re currently working on comprehensive analytics dashboards and automated suggestions for what to do in the event of various problems. The goal is to create a user interface that allows a production engineer who knows nothing about AI to solve problems quickly and easily.

I think the whole thing was very much driven by a pain point. The industry has a lot of pain points - companies need to become more efficient, they need to become more sustainable, they need to increase safety, and they’re struggling to find skilled labour. So there’s a lot of pressure on the industry, and they can’t maintain the status quo. So we came along and said this new technology can solve a problem you couldn’t solve before. We’ve seen a great response. I think, at the end of the day, it makes no difference to the customer whether it’s AI or not. What matters to them is that you can solve a critical pain point. How you do it is less important.

One of our investors suggested LEXR to me, because we had a lot of legal work in general. We also had investments from Silicon Valley investors in the US, so we needed a US holding. We had American lawyers as well, but LEXR handled everything for us on the Swiss side - from the founding of the company to licence agreements, service agreements, and other contracts. Because we’re in the manufacturing industry, where companies tend to be highly consolidated, we often deal with large companies. For example, we’re now working with Fortune 500 companies in the US, where legal issues are very critical and contracts are scrutinised down to the last word. LEXR has been a great help here, especially because Thomas and his team have always been very responsive. I usually had feedback within 2-3 days and was able to sit down with the client’s legal department very quickly.

Manufacturing isn’t the easiest industry to get into or do anything in, but it’s very rewarding, especially from a technological perspective. The impact you can have in this space is huge. My advice is to understand that you shouldn’t try to take a technology and find a use case for it. It’s better to find a pain point through exchange, and then invest in finding the technology to solve that pain point. So you should try to understand what pain point companies would actually invest in, how the problem is measured today, how it’s evaluated, and other such questions, in order to ultimately arrive at a solution.

The opportunities in manufacturing are huge. The industry loses about a trillion dollars today due to unplanned downtime - that’s a lot, and it’s a great opportunity to bring technology into the industry. In Europe, the shortage of skilled workers is a huge problem, because experience is being lost. There’s also great potential to find a way to retain this experience through technology. I think this is very important for the long-term future.

Book a non-binding discovery call