Editors Note:  This article was written by our friends at InData Labs and it will take you longer than 150 seconds. But it’s worth it.

It was through a helpful compilation by IT entrepreneur & investor, Aleksey Melnichek that we can see the expansion of the Belarusian AI map. According to the entrepreneur, the map will help to explain which industries AI works in and what technologies they use. Alexei hopes that with the help of experts and readers the list of companies will be expanded, and the map will be refined.

This list includes 60 companies that work in 10 different areas and there are already a number of world’s largest players like EPAM, Yandex, Viber, and young start-ups among them.

On the AI landscape there are companies that deal with “traditional” machine learning, natural language processing, and computer vision. and the list of specializations is also very wide. There are companies that use AI in the fields of healthcare (Flo, doc, Lung Passport), agriculture (OneSoil, Zoner.ag), retail, commercial manufacturing, finance, As a transport, and environmental protection.

A number of large companies have their R&Ds in Minsk. For example, Profitero, IHS Markit, WorkFusion, Yandex, Teqniksoft, and Viber. All of them, by the way, are residents of the HTP (High-tech Park).

With such an impressive portfolio and growing attention to the industry, we should expect that in the coming years we will hear new success stories of AI startups from Belarus; as it happened with AIMatter (passed under the control of Google), MSQRD (bought by Facebook), Zoner.ag, Juno (merged with Gett), and a large mobile developer Apalon, which came under the control of IAC Applications.

“AI is the new IT. At first, computers were introduced everywhere, now – artificial intelligence “, – said Alexander Chekan, venture partner of the Haxus fund, supporting science-driven AI initiatives. As a result, dev.by asked Belarus AI landscape participants how they estimate the prospects of the global and Belarusian AI.

EPAM: “Most of what AI means is used in our projects”

Most of what is usually meant by artificial intelligence is used in EPAM’s projects: natural language processing, computer vision, machine learning, neural networks, etc. ”I talk about trends based on what I see at work,” says Ivan Kravchenko, head of Data Science group at EPAM Belarus.

”Data Science, Artificial Intelligence, Machine Learning are very popular terms that penetrate deeper into various industries. For example, traditionally such tasks are common in marketing, but more and more tasks start coming from the manufacturing sector. Fancy abbreviation IoT (Internet of things) is increasingly growing by specific projects. Oil and gas companies have always paid much attention to data (at least geophysical), but the current market situation is actively pushing them to improve efficiency in all of their activities. As a result, their interest in optimization and forecasting models is growing”.

Ivan notes that he is personally more interested in the projects resulting in installation or production line work optimization rather than conversion growth, for example. At the same time, there are more projects that are aimed at automation of human work. That is also very interesting: “Strangely enough, in practice the success of such projects does not usually lead to massive employment terminations, people simply get more complicated tasks,” Ivan says.

The co-founder of EPAM, Leonid Lozner, predicts endless possibilities for AI and rapid progress in this area. “The trend is to try and use AI technologies in almost everything and try to make a startup out of it. Corporations are up-and-doing as well. ” Some of the AI achievements managed to get into our everyday life, for example, Alexa became an equal member of the family and “perfectly understands everyone”, and Spotify recommendations help to make pleasant musical discoveries.”

But is Belarus really aiming to be the capital of European artificial intelligence? In today’s world, Lozner calls caution on speaking about the benefits that are associated with only one country or region.

”Perhaps, only Silicon Valley can blatantly claim to be an exception. Although locals complain about the lack of sane cafes, where you can just eat normal food, not to mention the dramatic scarcity of women. Yet there are two major factors that are worth mentioning in Belarus. We have many highly skilled professionals, including people with an academic background, and several globally famous success stories of AI startups. And the people who made them happen are still at the helm, which means they are still capable of scaling their own success.”

Juno: “We are ordinary guys, we can stand on the shoulders of giants“

”Large companies – Facebook, Google, Microsoft are making large investments into their ecosystems. We are lucky, we can stand on the shoulders of giants,” says Arseniy Kravchenko, head of Data Science team in Juno.

His team uses the same set of open source solutions based on the Python-ecosystem, as most of the other data science teams do. Of course, there are some specifics and their own solutions, for example, geared to work with geodata. But such libraries as scikit-learn, xgboost, or tensorflow have almost become an industry standard.

It is difficult to speak about the long-term prospects of artificial intelligence in Belarus and throughout the world, in his opinion, it is not always possible to “even predict the future of the team in six months”.

”Democratization of data science solutions, knowledge, and software will lead to the fact that more projects will be able to use prebuilt machine learning algorithms. This won’t require hiring some brilliant guys with huge salaries. At the same time, application of some AI-algorithms is quite wide. From purely entertaining (all these applications with ridiculous effects) to conservative industries – heavy industry, medicine, agriculture”.

FriendlyData: “There are some high-class AI specialists in Belarus, but they are nowhere to be found”

FriendlyData deals with one of the traditional problems of AI, which is natural language processing. The stack is simple enough. The application is written in Ruby using other technologies (PostgreSQL, Redis, XSLT). The main business task that the team is solving is the transformation of the natural language into the language suitable for databases (SQL, for example). In order to do this, the startup has developed his own solution with an approach that is based on the principles of formal grammar.

Talking about AI perspectives, co-founder and CTO of FriendlyData Alexander Zaitsev focuses on two areas. The first one is natural language processing. Natural language user interfaces will be entering our lives more and more, replacing the classic search forms. For example, a bot that replaces a consultant in an online store will become more and more human-like and intelligent. Voice control of your home, telephone, and your car will become familiar and convenient. The second area is self-driving cars, which will enter our lives “relatively soon.”

In Belarus, it is quite difficult to expand the team that deals with AI, Alexander believes.

”There are some high-class specialists, but they are nowhere to be found. And constant research also requires considerable investments. Perhaps, in order to significantly improve the situation, it is necessary to reform our education system and make it more flexible meeting today’s requirements. We also should engage large companies like Google and Facebook in order for their R&Ds to appear on Belarus AI landscape”.

InData Labs: “Only a small part of enthusiasts are engaged in serious research on creating fundamentally new architectures of neural networks”

InData Labs is a data science & AI consulting company that builds a team and organizes staff training in such a way so that to be able to solve a wide range of problems in the field of Data Science and Big Data. In order to stay in trend, InData Labs invests in R&D. This year the company became the organizer of the Big Data Week in Minsk and the conference on artificial intelligence AI Day. The interest towards the event was much higher than the organizers expected. The registration was closed in two days after the conference was announced.

”For each and every project, we select the most appropriate technologies, based on the task, client’s needs and the amount of data available. Those can be both classical methods of machine learning and deep learning technologies, which have proved themselves in the tasks of image processing, text analysis, and time series forecasting,” says InData Labs CEO Ilya Kirillov.

We already have a symbiotic relationship with cars, mentions Ilya. It’s hard to imagine how the world will change when neural networks become even smarter. He is confident that people will have to learn to trust the machines and their decisions, without this, autopilot cars or AI in medicine, for example, will be useless.

Application of neural networks in various industries has recently increased. But most of such solutions are build using already existing mathematical approaches that can solve a small class of problems. Only a small part of enthusiasts are engaged in serious research on the creation of fundamentally new architectures of neural networks, strong and weak artificial intelligence.

Healthy Networks: “Belarusian school of advanced mathematics and statistics is pretty good, which is beneficial for us”

MedTech startup Healthy Networks is creating a product for the diagnosis of respiratory diseases based on neural networks. Recently, the project received $ 100,000 of investment from Spacemind Capital.

Lung Passport is a mobile application with an electronic stethoscope for monitoring and early detection of diseased changes in the sounds of lungs. In order to solve the problem, the team uses their own solutions and prebuilt ones. The startup hopes to eventually create a really big and unique software and hardware system.

”In Lung Passport, we use the Microsoft infrastructure, their new machine learning tool – CNTK. It is still trying to catch up with Tensorflow, but it is developed very quickly, and besides it is well compatible with the rest of the Microsoft stack, in particular with Azure, – says co-founder and CTO Aleksey Karankevich. – For our digital signal processing research and data preparation, we use open source software in Python. It is not possible to ignore ready-made solutions in modern realities”.

Aleksey associates the AI boom with the fact that the amount of data generated by the world has tangibly grown during the last couple of years. It’s nice to see that Belarus is not an exception to this trend, Aleksey is impressed with such projects as NeoSound, OneSoil, VibroBox, and Exponenta.io.

”Experts from Belarus may often lack machine learning training, but there is a good school of advanced mathematics and statistics,” said Aleksey.

He believes that Belarus AI landscape and community will continue to form in a natural way – on hackathons, contests, seminars, but first of all in workplaces in outsourcing companies, as it was first with Android and iOS, then with big data and blockchain. After all, outsourcing companies won’t stay aside from the AI-related tasks. Large businesses can not always find a ready-made service for their data and business goals so they have to turn to custom software development services. I would say that it is the main trend for the development of Belarus AI landscape in the near future.

Cybergizer: “Talented AI-engineers used to work in large companies”

A small company Cybergizer with a team of 25 people is working on several AI projects as an outsource partner (for example, an educational application for children based on Adaptive Learning. It adapts to the rhythm and peculiarities of a child). Cybergizer team is also working on their own project in the field of robotics.

On the basis of its R&D, the company develops its product – Robatz Network. This is an IoT platform that allows low-level robot devices to communicate with each other, organize group interaction and decision-making process in a group. Important components of this process are object recognition and clustering.

”In previous years large companies in our country could hire all the AI-engineers to work on their projects. It slowed down the development of other projects and startups,” says CTO Anatoly Lyotych.

”Now there are more people and companies on the market. This situation provides more options both for companies and engineers and makes the speciality not so elite. So next, I think, we should expect the emergence of new AI-startups on our AI landscape and the growth of AI-departments of existing companies. I believe that this stage of growth will last for a couple of more years, and then it will stabilize”.