Igor Ilyunin: “The customer asked to assemble a team that he did not find in Silicon Valley”

“The Internet of Things” is the direction in which development is moving at an incredible speed, where even strategic plans have to be revised almost every year. Igor Ilyunin, the leader of the IoT practice of DataArt, told about what the leading edge of IT is doing, how the approaches to hiring and training engineers have changed, what prospects it opens for young professionals.

I. I .: About one and a half years ago, we realized that in the IoT area the wind starts to blow a little to the other side. Previously, customers were interested in connecting devices, knowledge of specific data exchange protocols between them, various types of network connections, and building infrastructure at the device level. But at this moment several customers came to us - quite large companies, and one of them was a manufacturer of computer equipment. This company was going to build its own IoT-cloud, which would serve all their customers, wanted to ensure the movement of data, independently process them, allowing the customers themselves to do customization. That is, create a platform for all occasions. At the same time, they voiced a list of modern technologies that engineers of their supplier should have, and at that time we found only a couple of people in the whole company who at least partially met these requirements. I must say that the customer himself said: “We need a team of 3-4 people who will work on our project - we did not find such in Silicon Valley”. We saw this as a great opportunity - then the concept of the internal Big Data Academy appeared.
')
- Since the team of necessary specialists could not be assembled in Silicon Valley, then it simply did not exist?

- Yes. The client then immediately explained: he understands that we cannot have experts who already own all the necessary technologies, but he knows that we have competent engineers who are able and willing to learn. After that we assembled a team of three people. One of them was assigned a solution-consultant, at that time he was the most experienced, knew the Big Data landscape and could communicate with the client in the same language. Then for the first six months, the whole team was engaged in studying new technologies. Based on the knowledge gained, they did R & D and benchmarks, compared and tested how the system works. In principle, at that time we have assembled the most suitable engineers out of everyone we could find inside the company for the project.

Another client came with a similar problem. Their company makes smart beds: they put a thin mattress in the bed, which has sensors built in to the device that transmits all the information to the cloud. This allows you to track sleep patterns, and in hospitals to collect various data on the condition of patients - the accuracy of some sensors, even the heart rate allows you to track. The customer at the time of our acquaintance had about ten thousand devices, and even with this number of data processing systems, they barely coped with the amount of incoming information. In this case, the business had plans to reach the level of a million, and then in five million devices. They came to us after looking at our expertise in our own DeviceHive platform. But just after we began to have clients with such large projects, we realized that there was a lot to change in DeviceHive itself. Indeed, looking at your own plans for a year ago in IoT usually comes with a big smile.

The customer liked our concept, and we understood that in order to scale the platforms, advise the client, use the necessary resources, we need to learn how to train people. Now it's funny to remember how we tried to hire a person who owns the necessary technologies from the market. The vacancy then hung for four weeks without a single response, or rather, only a couple of courageous juniors responded to it, who, of course, knew nothing of what was said. We found ourselves in a stalemate: it was very interesting for us to develop technological expertise, but it was impossible to do this without engineers with the necessary qualifications. Then it became clear that people need to be trained independently.

- To teach all the technologies needed in specific projects of the company?

- Big Data is a huge collection of the most modern tools and frameworks. We decided to concentrate on some of them, because if you understand how 3-4 modern data processing or building scalable systems work, it’s much easier to get acquainted with the rest.

We took teams that worked on DeviceHive and the aforementioned commercial projects, and did a small course of lectures on Big Data: the main trends and tools. Now this course consists of six hours of video materials and practical tasks; it is located on the internal educational portal. A rather large group of people who periodically discuss market trends has formed around its preparation, new products - from Amazon to Microsoft - have developed their own technological get-together.

- The course can be considered successful? Those who listened to it, understand the new technologies?

- While we were recording the course, we gathered a small test audience of seven or eight students to understand how understandable the presentation of the material will be. As a result, they all got into projects that require quite serious technological knowledge. Some of them are already engaged in consulting in projects related to data processing and building scalable systems, others in engineering, while everyone is helping with sales.

- Do you plan to post courses in open access?

- In October, we are going to launch a test course for students. I see a trend - if before people were interested in learning Java, C ++, etc., now they are more and more attracted to solving specific complex problems. I myself came to this only in the process of work, now students already at the institute understand that knowledge of approaches to solving complex problems is more important than mastering a certain programming language. Let's try to conduct an introductory course on IoT and its combination with Big Data, let's look at the response. It would be interesting to establish contacts with a motivated group of students who are ready not only to come to us, but also to gather audiences who are interested in such tasks. We are planning to do a test run in Kharkov, where we have a fairly large group of participants from the Big Data Academy. If the output is all good, we will continue to work in other cities.

Now we understand where to send employees at the level of middle or senior, but it’s not easy to invite young people to your place. At the entrance there are few people, and we want to slightly stir up the students, to see who is really interested in advanced research.

- How high are the initial requirements for those whom IoT-project teams want to see at home?

- Recently, we reviewed the approach to inviting projects: if previously the center had a list of technologies and programming languages, now the desire to solve specific customer problems, the ability to switch from language to language, the willingness to learn new languages, frameworks and tools, the engineering approach to whole When we communicate with people, we look at how willing they are to learn, not only deepening, but also expanding their knowledge. Now in our projects, judging by the market in which we operate, this trend will continue for quite a long time.

- Just a year and a half ago, there was a breakthrough in understanding the prospects of the industry, and completely new people have already taken serious positions in projects. So, learning at IoT gives you a fast growth?

- And there is. The point is in the approach to evaluating specialists: for example, the market defines a person as middle developer, but we see that he is not shy about asking questions to the client, offers solutions to the problem and, in the end, gets the result - for us it looks more like senior. He may not know some of the features and subtleties of the language or technology, but at the same time do specific things right.

- It sounds attractive to anyone who does not know the subtleties of programming languages.

- Usually, it quickly becomes clear what a person is capable of. In principle, a good opportunity to get the necessary knowledge gives participation in R & D projects, one of which is our DeviceHive. In addition, projects arise at the sales stage, or when a customer is looking for a new fashionable technology that will be combined with the solution that he himself initially proposes. We love to attract new people to this kind of R & D. Naturally, they work together with consultants and more senior developers, whom we already trust from experience. And candidates who are not ready to improve usually drop out very quickly.

- Rapid changes in technology and the market as a whole have a strong impact on sales organization? Should I have a lot of research projects in IoT?

- Yes, over the past year we have had a lot of such projects, it is on R & D that a huge amount of manpower and resources is spent. And even if the project does not shoot, for us it still becomes an achievement, because we get experience that we can use in further sales.

- Does the structure of the R & D projects themselves change?

- If in the first R & D, which we did for the client, there were a lot of people involved, because, firstly, they were just interested, and secondly, we had little experience, then as it accumulates, it becomes easier to manage resources. Now we can put on R & D already for 1-2 people. Everyone knows where to discuss issues and problems, R & D closes much faster, and we have a very strong pre-sales team that can not only conduct research, but also draw conclusions from them. On average, our projects are designed for 1.5-2 years, some must meet in six months, but usually they are not associated with newfangled technologies.

- How are things now with a commercial project that prompted the idea of ​​Big Data Academy?

- Our team has made great progress. We did load testing, learned how to make the infrastructure linearly scalable. We can withstand a load of 100 thousand devices, for example, on a cluster of two virtual machines. We checked on four to work out 200 thousand, eight - 400 thousand and were able to demonstrate to the client linear scalability - he was all arranged. So now, if there is a specific task, we can easily bring the work of the system in line with the scale of the customer's business.