Sergey Nikolaev and Mikhail Gusev told in an interview for the TASS project "Conversations with Ivan Survillo" about the creation of a domestic digital process control system
Sergey Nikolaev, Mikhail Gusev and Sergey Belov founded CyberPhysics, a company born in the laboratories of Skoltech, creates a domestic digital process control system. Its implementation makes it possible to reduce the number of accidents, shutdowns and downtime of equipment, save resources and improve the quality and output of products. The hybrid modeling technology underlying this solution makes it possible to make forecasts both based on the technical condition of the equipment and on the basis of physical and mathematical models.
Sergey Nikolaev and Mikhail Gusev told in an interview for the TASS project "Conversations with Ivan Survillo" about the activities and prospects of CyberPhysics.
— How would you tell a five-year-old child what you do?
Sergey: Yes, I have a five-year-old child. We discussed work with him recently, and he likes what I do. Because he also says he works for CyberPhysics. More precisely, he says that he has a company called Super-Cyberphysics, which is even better.
In fact, this is not quite true, because we do not work with trains directly. But it's true about turbines. What we do allows us to prevent accidents and destruction.
To be more serious, we are making software that monitors the condition of large, expensive, critical equipment and allows us to find small incipient signs in it that something is going wrong. Malfunctions or defects that have not yet happened and are only appearing. We predict them even before they develop to a critical level.
The second point concerns optimization. There is less heat of passion, but this is also important. Mish, can you tell me roughly?
Michael: Yes, in terms of optimization, our software helps to reduce gas consumption in a large industrial furnace. The cost of money for this gas is 17-20 million rubles per month, and our programs allow you to save gas, give recommendations on the control parameters of an industrial unit and save about 7-10%.
Sergei: All this is achieved through the same mathematical models. To put it pathetically, digital counterparts of processes and machines that calculate in real time how it should be, are compared with how it is, and the main methodology is based on this difference. We analyze the differences — we call them "differences" — and indicate which internal parameters affect the differences the most.
— I want an example.
Sergey:A good example is with a gas stove, but it has already been. Therefore, I will give another example with the chemical industry.
There is the production of a certain chemical, for example, chloroform — a fairly useful chemical in industry. There are certain parameters that control and regulate the performance of this process. Their in-house values float in a fairly wide range. They do not take into account specific changing conditions — they are set once, so they exist.
Our software analyzes much narrower and more accurately. He looks when the process starts to go wrong, sees it by performance and understands what affects it: the temperature of the reactor, the pressure at the inlet to the compressor, and so on. There may be dozens of such reasons... Next, our software determines which specific control parameters need to be tweaked so that the output indicators change to the required values. These are usually the parameters that can be changed from the operator console.
The installation operator sees that our system is signaling something: "Yeah, it's like the factory system tells me that everything is fine, but this more subtle system tells me to increase the temperature by five degrees." Then he takes it, increases the temperature by five degrees, and productivity increases. It increases by percentages, and interest amounts to tens and hundreds of millions of rubles in monthly or annual terms.
— And why do you need an operator?
Sergey:That's a good question. The next stage in the development of our system and AI class systems in general is just work without human participation, when the system itself fully regulates production. Everything will come to this. Our system will eventually evolve into such a system, but there are still people. So far, real intelligence controls artificial intelligence in most enterprises.
In fact, there are no obstacles to the opposite, except for the level of technical equipment, business processes and trust in the systems on the part of management.
We always send an engineer to technical negotiations who can speak the same language with other engineers. Chemists need a chemist, turbinists need a turbinist, and so on. This is very important, because the "friend-foe" marker is triggered. If you are a very cool specialist in metallurgy, but you came to the chemists, they will not consider you as a person.
The human factor is very important to take into account, so we have an engineering-oriented company. Engineers work for engineers — we are trying to create this fleur around ourselves. If we had only specialists in mathematical modeling, Data Science and AI, we would never have been able to find a common language with production workers. Many colleagues-competitors make this mistake and break up not because they have bad math, but because they cannot convey their bright ideas to production workers.
MICHAEL:I will add: now engineers in Russia are undervalued by salaries. We are a little ahead of the market here and we are making our contribution so that engineers can do interesting tasks, understanding only their engineering specialty, can solve real production tasks and earn more money than the market currently gives.
— New IT specialists?
Mikhail: Yes. Engineering elite.
Sergey:Or digital engineers, something like that...
These same engineers will be setting up complex automation and production management systems in ten years. They will have to know all this math and not be disconnected from how the hardware works at the factory. Whatever our digital world may be, after all, our material values are produced in factories with pieces of iron. They're just smarter now.
— By the way, what are the hardware data based on in your software?
Sergei: We collect data from sensors. They are always installed if we are talking about more or less modern equipment. We don't work with the very ancient one ourselves — first we need to put sensors there, they have to measure everything in continuous mode for a year, and then automation, digital doubles and everything else.
Maybe Soviet ancient equipment, but with sensors. The same are our favorite turbines or rolling mills, chemical lines made in the 70s, but retrofitted with automation tools. They are quite comparable to modern ones: they have less efficiency, but they still work the same way. It's about the layers: first the hardware, then automation, and then we.
We analyze data from sensors using mathematical models, and sometimes we invest physics more, so CyberPhysics.
— For example?
Sergey:For example, if no one knows about the equipment and it has no characteristics, then only sensors will help. However, if the equipment is known, there is access to the manufacturer, you can request design and factory specifications from him, and we can integrate a physical model. That is, where there is no sensor, something can be modeled using an equation — for example, temperature.
MICHAEL:Just yesterday I had a meeting. One of our areas is temperature modeling during the continuous casting of steel. Defects in such a process can be determined only after the products have passed the entire production cycle, when there are already ready slabs (thick steel billets of rectangular cross-section).
We have made a model that allows us to evaluate defects directly in the process when hot metal is being poured. Naturally, nothing can be measured there, there are no sensors, there is just liquid hot metal.
The essence of physical and mathematical modeling is that we describe the process with an equation and pull out so-called virtual sensors from these equations. They show us a parameter that cannot be measured directly, for example, the "damage" of metal in a continuously cast workpiece.
Yesterday we demonstrated to the customer the relationship between the "damage" and the resulting defects of the workpieces — indeed, from the analysis of these virtual sensors it was clearly visible how to reduce the number of defects, the customer was very happy.
Sergey:These are places where it is impossible to put a sensor for objective reasons, or sometimes it is an opportunity to simulate those situations that did not exist in real history: drive your equipment virtually into a bad situation to see how it will behave, how to configure algorithms. This is very important, especially for new equipment, when history has not accumulated yet.
Again, unlike our competitors-colleagues, we can work with manufacturers. The manufacturer can use our software from the first day to service the conditional turbine according to the condition, understand what is happening in it, and analyze it. Even with a minimal amount of accumulated data, there is already a certain primary model that describes more or less adequately. And then, with the help of data, it is clarified.
— Aren't you thinking of cooperating with the manufacturer?
Sergey:Of course, that's what we thought from the very beginning. Initially, we had two global directions: operators and manufacturers. So far, we have most of our revenue from the manufacturer. First of all, our Russian engine-building corporation, which produces gas turbine engines. We have been working with them for a long time and for the future.
We have certain industries that are now important within the framework of technological sovereignty, we are really working with them. Another example is the "Power Machines" in St. Petersburg. In Russia, there are no large gas turbine power plants of their own, they are now the first to make them, before that there were Siemens. But Siemens, as you know, have left, so we are already interacting with the "Power Machines" in the same direction, so that they eventually supply the system assembled with our software.
About import substitution, shifts of lithospheric plates and the sale of the company
— How does everything that happens affect you in general?
MICHAEL:On the one hand, many companies that were our competitors have really left the Russian market. We began to receive more applications for the replacement of foreign software in terms of, for example, predictive analytics. We mainly work with companies of a continuous cycle: oil and gas, mining, metallurgy, energy. These industries form the basis of Russia's export potential.
On the other hand, the market has some inertia, that is, we cannot say that we will be able to replace everything at once. Enterprises should include us in programs, find financing... From the point of view of turnover, there is not much growth right now. We expect it by the end of the year. In general, the situation is going well for us, but maybe the CEO will add something here?
Sergey:The situation is promising and promising for us, but at the same time it is very difficult at the moment.
There are large shifts of lithospheric plates. The giants are leaving — we cannot be comparable in terms of volumes and capabilities with AVEVA or Schneider Electric, they were closing a huge niche. Small companies like us can close some specific things, but a large customer needs to put a stove and close the hole entirely. If you put a small stone, it will fall, and there will be no sense.
In this regard, on the one hand, there are great opportunities, on the other — our customers tell us: can you do it? And we say no, because investments in this type of product are hundreds of millions of dollars. I have to put a patchwork quilt.
Of course, we claim that we are better than them, and we are, but just imagine: there is a company with a staff of a thousand engineers, and there is a company with a staff of ten. I'd take the one with a thousand myself. Why would I take the risk?
— If you dream, in 15-20 years — what are you like?
Sergey:We're doing something else—at least I'm sure of it. Our company and technology should then be integrated into a larger ecosystem, like Schneider Electric, like Siemens. I personally don't think as CEO right now that we can be at the head of this. Most likely, we will do so in three years already. But that's how I see myself, Misha doesn't know what he'll say.
— Does the development director have completely different plans?
MICHAEL:We are now faced with real business systems, we understand who works how, how to move projects... I don't want to spend this skill on one thing, it seems that it can be extended to several directions.
My idea is somewhat similar to Sergey's idea - to develop several technology businesses. We already understand and understand very much what customers really need, what will be in demand in ten years. Now the main task is to develop the product we are making to a super level, to collect the maximum number of errors so that the next projects can be grown not in three years, but, for example, in two.
Sergey:I fully agree — indeed, the mistakes and experience that you gain while developing the company are, as a rule, common, and then you can help other guys.
I would have helped myself three years ago, I would have said — don't do this, it's useless, don't waste time on this... From now on, our technology should live and develop not as a startup, but as part of a large technology platform.
— Usually everyone says they want to create such a platform themselves.
Sergey:Absolutely all startups that do what we do end up being sold for hundreds of millions of dollars to larger corporations that close large niches. Either startups go for an IPO, but this is a much rarer scenario.
Making a company for an IPO is a long time, 10-15 years, I haven't seen any other examples yet. In my opinion, if there is an entrepreneurial streak, it is better to make a good product, sell it, get funds and trust loans to do something else. I have such a position. By the way, the same Elon Musk did so: he sold his first companies a long time ago and did something bigger. If he was still developing a payment system, then he would be nobody objectively. In this regard, we borrow a little idea from him.
About mistakes, subtleties of negotiations, zealous developers and the triumvirate
— You were talking about the number of errors that are being stuffed. What was the biggest one already?
Sergey:There was a lot! I don't even know where to start...
Typical startup mistakes that we have made are to go into complex development without calculating the consequences. Our first version of the software was very beautiful, but absolutely uninhabitable. To support it, we needed three times as many developers, and we didn't have money for them then.
It's a vicious circle that you can't break.
In general, it is not necessary to make a super-large, beautiful, complex product from the very beginning. We started doing this and cut down some of the money — thank God, we had it from an investor, I hope Fabio (Fabio Massimo Caccatori is a serial entrepreneur and the first investor of CyberPhysics — approx. TASS) will not read this interview.
The second mistake, which is very common, is to hire beginners for key positions with the hope that they will learn and everything will be fine.
The third mistake is connected with the second: to assume that employees are your friends—entrepreneurs, treat them the same way and demand from them as from yourself. But in fact, the employee does not care about your business at all — and that's right. He should not care, he should just do his job well.
For an employee, your startup is just a place of work. It can be pleasant, cool, with an atmosphere, with fun, with holidays, with events, but it's still a job, and he will change it later. And for you, it's a matter of life. It is necessary to divide it. At first I didn't share, for me everyone was like Misha. But Misha is a founder, and you need to ask him as a founder, and you need to ask an employee clearly for his function.
MICHAEL:I will add: an obvious mistake that definitely occurs at all startups — when we opened, we were expecting that a number of projects were about to go, we hired some people who fit these projects. And then reality kicked in, and no projects went ahead. It is necessary to understand super clearly that your project will definitely go, and hire people when it has gone. Otherwise, a lot of money and time is really burned in vain.
Another of the mistakes — their consequences have not yet been completely eliminated, by the way — is to promise anything at the first stages, just to get a contract. Then, when it comes time to pay the bills, all this may not pay off, and the project may turn out to be really unprofitable, problematic and bad. Precisely because you initially promised something that most likely will not be done. And you promised because you really wanted to get a contract. The clear motivation is to get something as soon as possible, but this is the mistake of a beginner, a young, insecure person.
Sergey:It seems to me that if you have already lived, you'd better hold more negotiations, but you will find a customer with whom you can justify what you will actually do. Then it is much more likely that you will get a good case. Without experience it is difficult to realize. Before that, we were engaged in science with Misha, there is no such thing there. You do a scientific study, hand it in, and that's it.
In general, there are extremes in how customers perceive us. The first is super—extreme skepticism: nothing works, everything is a complete mess, you don't know anything, everything works for us, nothing breaks for us. The second extreme is the belief in a miracle: the belief that we are doing magic. For example, a client says: we have equipment, there are no sensors. You can do something with the help of a physical model, right? You promise us something very cool — so you can do something, am I right? I think to myself: it seems like you have a technical education, what exactly is "something"? Conjure with a wand?
— Do you often argue with each other?
Sergey:We have a triumvirate, like in ancient Rome — by the way, one of them finished badly there, his name was Crassus. He loved gold, he went to the Parthians... But no, we have another triumvirate.
We have me as CEO, Mikhail as Development Director, and Sergey Belov as Technical Director, also a co—founder. He is responsible for the architecture of the product and its development.
We have such a position among ourselves: Sergey has his own vision, and it is a product vision. He is responsible for the development and has a bad attitude to any improvements or deviations from the product that lead away. But sometimes such things are needed because the customer wants it.
Usually, Mikhail brings Sergey information about improvements and "features", because he communicates a lot with customers. Here disputes and discussions begin — is it really necessary or did Misha come up with it, because he's funny and cool and he won't have to do it, and Sergey will have to figure out how to implement it, implement it into the product, allocate resources for development, and everyone is already loaded to the eyeballs. I seem to be standing in the middle, trying to find consensus.
Developers love to develop, and if they are not controlled, they will develop something. It's beautiful, but then you come to the customer with it, and he doesn't understand what it is. At the other extreme, if you only work with customers, you can throw a bunch of ideas, do not have time to do anything, and nothing will work. Therefore, it is important to find a middle ground in the "confrontation" of salespeople and developers.
Of course, I also occasionally have a desire to finish something, or even dissatisfaction with the fact that something else is not working. But we have never had misunderstandings, resentments. Each of us is aware that our controversy is part of the workflow.
MICHAEL:Yes, I'll note it on my own. I also have an extreme — if I explain, I immediately remember a bunch of "jambs", I report them all calmly, carefully, methodically, and the guys just have a head from this smokes. On the amount of what needs to be done. But they can't do it right now because it takes time and resources are limited.
— What is the main thrill for you in all this?
Sergey:It's easy for me to formulate, I even answered this question recently — we were sitting, having dinner with our first investor, Italian Fabio. He lives in Russia, in Moscow, and deliberately, he even got citizenship, an ass.
Considering that we are a very small company and the customers are very large, this is a special challenge. Every time I win this challenge, it's a special thrill for me.
The second thrill is when the project really turns out. But this is not my merit, but those who do. When I see that the result satisfies the customer, everyone is happy, satisfied, respect shines in their eyes.
MICHAEL:I personally get high from the process. You find an interesting thing, it can be implemented into software — and here you are. Even if not now, even in three months, but it can work! That is, you really bring value, you create a product. This is probably the greatest pleasure I get from all the activities.
— At the end, let everyone name the project they are most proud of.
Sergey:Increase in the productivity of the chemical production line by 10%. Our system understands when production is not working optimally and suggests how to improve it. All this gives a sufficiently powerful effect. We cannot name the customer, but this is a chemical enterprise, quite large.
MICHAEL:My case also relates to production optimization, but it is the "physics" part of our CyberPhysics name that works there. We made a physical and mathematical model that really helped to reduce the mixing time of steel in the vacuum cleaner by half. The customer tested this on a real metallurgical ladle, where steel is brewed, and the test results showed exactly the same things as in the model.
Sergey:We can work in the field of data-science, and in the field of physics, and at the junction of both.
