Translated by
2020/03/07 10:51:02

What clever women are attracted by the sphere of Big Data and artificial intelligence with. 9 stories from Russia

The women conducting researches in the field of Big Data, artificial intelligence and machine learning remain in minority so far, but at the same time contribute significantly to research activity. They participate in interesting projects which results often find important practical application. TAdviser communicated to several such heroines and found out that it attracts them in this sphere and in what projects they are involved.


Despite the policy of the large technology companies which seriously attended in recent years to attraction of bigger number of women in the industry the IT industry remains mainly men's. And in the areas connected with data, such as Data Science, artificial intelligence, machine learning and others the gender gap often is even more, than "on average in hospital", especially in the academic environment.

It is characteristic also of the countries which are technology leaders. For example, in the USA, according to the research of recruiting company Burtch Works published in 2018, only 15% of specialists in the theory and methods of data analysis – women. And the research conducted by AI Now Institute in 2019 showed that less than 20% of researchers in the field of the artificial intelligence (AI) submitting applications for participation in prestigious conferences in this sphere – a female and only about a quarter of the students studying AI at the Stanford university – girls.

Work with data - it not only is interesting, but also it is fashionable (a photo -

Similar situation and in Britain. There the recruiting company Datatech Analytics in 2019 found out that women hold about a quarter of all positions connected with Data Science.

Estimates of a number of respondents of TAdviser of employees of research teams in universities indicate that Russia in this respect, one may say, in a trend. They estimate a share of the women occupied in researches in the field of data, on average in 15% though in certain areas call also higher percent – to 35-40%. For example, in the field of research and development in a scope of the artificial intelligence (AI) for the problems connected with environmental protection and agriculture.

At the same time some respondents note a tendency to increase in a share of women in the research environment in the field of data, AI and machine learning.

It is considered that such work is complex for women. I do not agree with it: at us is more and more the female research teams working in scientific and analytical community. I think that it was affected by the high level of education in Russia in natural and technical science. Everything in the world begins with the idea as still the Ancient Greek philosopher Platon showed. Also he argued on the capacity of the society capable to generate and realize these ideas. The person can develop, will always pull the person to novel areas: and in external space, and in internal intellectual space, - Tatyana Podladchikova, the senior teacher of the Space center of Skoltech, Candidate of Technical Sciences, the applied mathematician notes.

For this reason, she thinks, now there are a lot of colleagues of women with high potential to create worthy and interesting ideas.

The Russian researchers in the field of Big Data, AI and machine learning participate in interesting projects which results often find important practical application. For example, they are applied to forecasting of space weather, reflection of floods, detection of zones of defeat of a weed of the Cow-parsnip, dangerous to the person, Sosnovsky and many other. TAdviser collected stories of several such heroines.

Space weather

Tatyana Podladchikova, the senior teacher of the Space center of Skoltech, Candidate of Technical Sciences, the applied mathematician, the winner of the international medal of Alexander Chizhevsky in space weather and space climate (the Photo - Leonid Sorokin/Inc)

Tatyana Podladchikova is engaged in studying of the Sun and space weather using AI. Space weather requires permanent monitoring of activity of the Sun and space and is directed to development of operational services of forecasting and reduction of effects of the extreme space weather phenomena, she explains. However development and improvement of services of space weather requires in-depth studies in the field of solar and terrestrial physics, understanding of processes in the Sun and all subtleties of interaction between the Sun and Earth.

Methods of fast classification and the analysis of a flow of solar images on which we work in our laboratory together with the international colleagues allow to detect powerful emissions of solar weight and also to predict their arrival to Earth. In case of an alarm, as a rule, switch off the sensitive equipment of satellites which fly around Earth. Our developments and services for forecasting of solar activity are used for radiation level assessment at the flight altitude of airplanes, - Tatyana Podladchikova told TAdviser.

According to her, together with the European Space Agency development of new service of forecasting of radio of a flow from the Sun is also conducted that has great practical value for a time estimate of return of spacecrafts to Earth, correction of orbits of satellites, warning of collisions and modeling of space debris. And together with the university of Graz, Austria, goes work on creation of service of forecasting of a high-grade solar wind at Earth and the related geomagnetic storms.

Besides, within the big European project on creation of the 4-meter land solar telescope the laboratory works on algorithms of the images received from different ground stations for ensuring high-quality observations and detecting of flashes in the Sun. New achievements in the field of artificial intelligence allow to improve images of lower quality, received by telescopes of last generation on the basis of new high-quality these modern telescopes, Tatyana Podladchikova explains.

A series of continuous observations with the best quality will allow to understand more deeply key mechanisms of difficult physical processes in the Sun for a long time that in turn will allow to improve operational services of forecasting of space weather, she says.

"Smart" condition monitoring of heart

Data analysis methods which are developed in laboratory under the leadership of Tatyana Podladchikova are applied not only in space area.

Data analysis methods which we develop in our laboratory also are applied also to extraction of useful knowledge, control and forecasting for cross-disciplinary applications. The research project in Skoltech on creation of the wearable device with artificial intelligence for monitoring and the analysis of a status of a cardiovascular system developed into SENSE2BEAT startup, - explains Podladtchikov.

An ideologist and the project manager is the graduate student of the space center of Skoltech Natalya Glazkova. Darya Stepanova, the graduate of Skoltech, is responsible for a technical part of the project.

Main advantage of the developed device - in constant control for the cardiogram, Tatyana Podladchikova says. It allows to perform primary diagnostics of a number of cardiac arrhythmias which are difficult to recording by means of single short inspection on the stationary cardiograph. This device will be useful for astronauts – during the land trainings, flight and the subsequent rehabilitation, to athletes for monitoring of work of a cardiac muscle in extreme conditions and the corresponding adjustment of the plan of trainings. Also it would be useful to all people on Earth for significant improvement of quality and life expectancy.

Artificial intelligence for the oil and gas industry

Ekaterina Muravleva, senior research associate of the Center of hydrocarbon production of Skoltech

The main area of interests of Ekaterina Muravleva - mathematical modeling. Developments of scientific group of Dmitry Koroteyev in which it works are already implemented in several large oil and gas companies, reported in Skoltech.

We develop tools for decision-making at exploration and production of oil and gas which allow to reduce costs at technology transactions on fields and to objectively estimate the capacity of oil and gas fields, - Muravleva explained TAdviser.

Speaking about the importance of the researches which are conducted with its participation in the field of AI, Ekaterina Muravleva noted that progress of AI is generally connected with image identification and image processing, video, texts and a sound, and they only should find the place in the classical fields of calculus mathematics. Nevertheless, the potential of methods of machine and deep learning in problems of modeling is huge and involves a large number of researchers worldwide. They allow to lower the cost and time of calculations, and in certain cases the AI methods allow to increase significantly quality of modeling, Ekaterina Muravleva says.

Fight against pollution and weeds

Maria Pukalchik, the senior teacher of Skoltech, the Center for scientific and engineering computing technologies for tasks with data bulks (CDISE), Candidate of Biology

The scientific group in which Maria Pukalchik works is engaged in fundamental and applied developments, it is focused at works in the field of analytics of soils, plants and quality of the environment. The following developments can be referred to the most significant achievements, Maria Pukalchik considers.

In the field of agriculture, based on given about structure of soils and their agrochemical indicators, we are able to build models of normative productivity of crops and to reveal the most "important" indicators of soils for their productivity with a binding to any region of Russia with the permission to separate fields, - she tells.

Based on given about type of soils and pollution level by oil, the group also created the model based on AI which can foretell harm for a vegetable cover (phytotoxicity) that can accelerate considerably decision making process about need of recovery or a remediation of the contaminated territories for oil-extracting regions of the country. Results of this work will be published in article of the Q1 Ecotoxicology and Environmental safety magazine in the nearest future, Maria Pukalchik says.

Besides, AI technologies (convolution neural networks, Convolutional Neural Networks) on the single board computer placed onboard the unmanned aerial vehicle (UAV) in real time to reveal zones of defeat of a weed plant of the Cow-parsnip, dangerous to the person, Sosnovsky were implemented.

AI is not tired, and, unlike a human eye, reveals even those single plants which would be passed by the people watching the material which is finished shooting video from the drone without the AI system, - Maria Pukalchik explains.

Speaking about the importance of activity of scientific group, she noted that the favorable environment is a guarantee of health and wellbeing us and our children and also the guarantor of future development of the whole country. All of us daily see how the environmental problems connected with activity of mankind such as pollution of soils, air and rivers, climate changes, decrease in a biodiversity become aggravated.

At the same time, a set of the processes and the phenomena describing the environment and also high degree of uncertainty, variability and randomness it is possible to provide to them mathematically as a flow of the unstructured multiscale data arriving continuously in large volumes, so and many questions connected with environmental protection and agriculture it is possible and it is necessary to solve using AI and methods of machine learning, - Maria Pukalchik noted.

DeepPavlov and bot for Amazon

Диляра Баймурзина, исследователь Лаборатории нейронных систем и глубокого обучения MIPT

Primary activity of Dilyara Baymurzina in laboratory is devoted to natural languag processing (natural language processing).

We are engaged in development of open-source of DeepPavlov and dp-agent libraries for creation of chat-bots. From summer of 2017 to summer of 2019 I develop and I support the component of library which is responsible for text classification. Also in 2018 I participated in tender on Kaggle Toxic Comment Classification Challenge in the course of which we began experiments on evolutionary selection of parameters of models. After the end of tender in DeepPavlov library there was also an opportunity to select hyper parameters of models, and we began work on evolutionary selection of neural network architecture, - she tells.

The code with these developments is still not laid out in open source, but in the nearest future the project on selection of architecture will return to an active phase again, and, Dilyara Baymurzina hopes, results after all will be available to users of DeepPavlov library.

The reason of freezing of the project on selection of architecture she calls that last summer the command of MIPT DREAM was selected for participation in the tender Amazon Alexa Prize Grand Challenge 3. Therefore in August team members began to be engaged in development of a bot for Amazon all the working time.

We successfully underwent certification of a bot in October of last year, and in December our bot DREAM had an opportunity to communicate with users of the smart columns Alexa. At the moment there is a tender quarterfinal, and I hope, our command successfully will pass in a semi-final and will continue development of DREAM of a bot. For me participation in this tender was fantastically useful experiment on development of a bot in a produktiva as such work strongly differs from scientific activity, - Baymurzina says.

She hopes that experience of a command in the form of a ready bot which will be laid out in open-source on the end of tender will be useful in community of a natural language.

Anastasia Kravtsova, specialist with data of Laboratory of the neural systems and deep learning of MIPT

Anastasia Kravtsova works in the same laboratory, as Dilyara Baymurzina. It also consists in a command which is engaged in development of open-source of DeepPavlov library for creation of dialogue assistants and the analysis of the text.

We solve the problems connected with natural languag processing by creation and training of neural network models for the analysis of tonality, recognition of named entities, answers to questions, etc. From these components it is possible to collect a full-fledged dialogue system under your applied needs. Directly my work consists in collecting and the analysis of text data for creation and assessment of similar models, - Kravtsova explains.

Fighters against floods

Анна Калюжная, руководитель научного подразделения в Национальном центре когнитивных разработок ITMO university

Initially Anna Kalyuzhnaya specialized in modeling of hydrometeorological processes on the basis of hydrodynamical equations and multidimensional statistical analysis. However now thanks to the accumulated experience of participation in different applied projects its interests and scientific ideas stretch far beyond modeling of hydrometeorological processes.

The last several years I and my command we work on development of the data-driven methods of modeling for different applied tasks, we have scientific and R&D projects in this area, - told Kalyuzhnaya of TAdviser.

Projects include:

  • intellectual design of wave protection constructions in the port water area using evolutionary approaches (the method of identification of optimal structure of wave protection constructions is developed);
  • model of credit scoring on the basis of data on transactions (developed intellectual model of credit scoring in which risks assessment is made using creation of a behavioural profile on the basis of the history of bank transactions of the client).

At the same time, tells Kalyuzhnaya, and about modeling of natural processes do not forget.

In September, 2020 we begin set in the magistracy "Digital geotechnologies" devoted to technologies of the analysis of space geodata and geomodelling, - she tells.

Anna Kalyuzhnoy's team also participated in development and putting into operation of the system of prevention of floods in St. Petersburg created by ARIA company for Directorate of the Complex of protective constructions. They developed the "brain" of a dam or the knowledge-intensive modules which are responsible for improvement of quality of hydrodynamic forecasting of a wind-induced wave and decision making support system on the basis of the algorithms creating the optimal schedule for maneuvering by locks told TAdviser at the ITMO University. Result – tens of reflected floods more than in 8 years of successful work of a dam.

Data analysis of social networks

Ksenia Mukhina, research associate of the National center of cognitive developments of the ITMO University

Ksenia Mukhina is engaged in data analysis of social networks: generally this research of activity of users in certain locations.

On the basis of it I together with colleagues developed a method for extraction of events which allows to detect events with an accuracy of 77% and 97% completeness. Still I use data of social networks and Internet resources for creation of automatic walking tourist routes. Survey results showed that the routes prepared by our system are pleasant to people even more, than expert, - Ksenia Mukhina told TAdviser.

Who stands behind quality of search of "Yandex"

Екатерина Серажим, руководитель службы релевантности и лингвистики в поиске "Yandex"

Ekaterina Serazhim is responsible in "Yandex" for quality of search and heads service of relevance and linguistics. When the user sets to the search system a request, as the answer usually serves a set of links. Estimate manually or automatically as far as this set responds to your request, just and means to measure quality of search. Ekaterina Serazhim's team develops and implements new models and algorithms that search results corresponded to requests of users even better.

She participates in improvement of search algorithms since 2012 when she came to the company the analyst. Those years the search queries which arrived from different regions and relating to different subjects were processed by different algorithms. It did not allow to enter global modifications into the system — it was required to change each algorithm separately. Serazhim and her colleagues constructed the flexible and easily improved system in which the uniform algorithm processes all requests from users.

In 2015 Ekaterina Serazhim became the head of group of analytics of quality of search. She and its command took active part in preparation of two major updates of search — Palekh (2016) and "Korolyov" (2017). These updates were based on neural networks. Use of neuronets in "Palyokha" allowed to teach to find search in the Internet of the page which correspond to requests not only on a key word, but also on sense.

"Korolyov" was the next step — neuronets in this updating were used even more actively, including at the deepest stages of ranging (sorting of links).

The group of analytics under the leadership of Ekaterina Serazhim was engaged in ranking models for "Palyokha" and "Korolyov". Then she was engaged in development of quality of search on mobile devices: in particular, worked on model in which the websites are ranged taking into account their mobile adaptivity. Including thanks to it the share of search of "Yandex" considerably grew by Android OS and continues to grow, note in the company.

Method of machine learning which was pleasant to Netflix

Anna Veronika Dorogush, the head of group of ML systems in "Yandex"

Anna Veronika is engaged in development of machine learning in "Yandex". It one of developers of a method of machine learning CatBoost and head of this direction. It is developed in Yandex for solving of tasks of ranging, prediction and creation of recommendations.

CatBoost helps to work with different data types: numerical, categorial, text. On the basis of these data it builds model for prediction of results of events which the algorithm did not see. The method allows to analyze more difficult, diverse data and to consider bigger quantity of factors. This algorithm is based on technology of a gradient busting which allows gradually, with each further step to improve results of the previous steps.

CatBoost was uploaded publicly in the summer of 2017. Since then it is used in the different companies worldwide, including Netflix, Careem taxi, CloudFlare, Aviasales and many others. Also it is used by scientific community in medicine and physics.

Now Anna Veronika Dorogush and her command are engaged in development of library, adding of support of new data types in it, improvement of an algorithm, integration into other platforms. Besides, Anna Veronika's team is engaged in creation and development of other libraries of machine learning which are used in "Yandex".

What attracts women in this sphere

The aspiration to be in harmony with the world around attracts interest of the person in understanding of an essence of the observed phenomena, their patterns and anticipation of further succession of events, Tatyana Podladchikova from the Space center of Skoltech says. Observation and an experiment are the cornerstone of knowledge of the world around. Every day we obtain a large number of the valuable data. However the most part of information is lost as we cannot process data rather effectively, she notes.

In recent years methods of artificial intelligence achieved remarkable results. And it is true success when we can receive new useful knowledge from the data leading to understanding of an essence of the observed phenomena, control and forecasting of future succession of events. And also to come to the reliable decisions made on the basis of the received results, - explains Podladtchikov the interest in this sphere.

The flow of text and other data grows every day, and someone needs to process and take these data from them something useful, Anastasia Kravtsova from Laboratory of the neural systems and deep learning of MIPT says.

Besides, demand for dialogue assistants and automation of communication in general when addresses are processed without participation of the living person grows so perspective make life of people easier should inspire, - she notes.

When Anna Kalyuzhnaya, nowadays the head of scientific division in the National center of cognitive developments of the ITMO University, ended training in university, she wanted to find work which, on the one hand, would offer it a call — assumed the regular solution of interesting challenges, and, with another, would become that place where it could bring benefit to society. She consulted to the research supervisor, and he told her about Alexander Valeryevich Bukhanovsky, the director of the National center of cognitive developments, and those scientific projects which it develops. Thus, Kalyuzhnaya began the scientific way.

Ekaterina Serazhim from "Yandex" became interested in machine learning still being a student at the university. At that time it was not such noticeable direction yet, she says, and in Tower on applied mathematics there was only a two-three of rates on this subject.

One of them me so interested that I was engaged in scientific work in this direction. Already then, in student's years, I was struck that, how abruptly algorithms of machine learning can solve practical problems like forecasting of weather or object search on the image. It seemed some magic. I decided to deepen the knowledge and arrived to study the School of Data Analysis (SDA). After its termination the choice to go to work to "Yandex" became for me obvious. It is possible to be engaged in something like application studies here — to think out something and to experiment constantly, - Serazhim explains.

According to Ksenia Mukhina from the National center of cognitive developments of the ITMO University, now unique time when there are a lot of data on what occurs around us, became available to researches.

People not only invented special sensors to take information on the world around, but also thought up social networks on which it is possible to understand that it is interesting to people and that attracts them. And if it is correct to set the task, then it is possible to find the answer almost to any question, - Mukhina explains.

Ekaterina Muravleva from the Center of hydrocarbon production of Skoltech tells that it with honors graduated mekhmat from MSU and defended the dissertation on calculus mathematics. Can seem that a lot of things that teach on a mekhmata to - it is abstract things, but actually, at AI and classical mathematics many general points. For example, topological approaches are rather actively applied, she says.

It is important for me that results of my researches were demanded that cannot be told about many theoretical researches. The AI methods became such area on a joint of the theory and practice actually. Now it is very difficult to be engaged in something, connected with computing and numerical methods and not to be interested in AI, - Ekaterina Muravleva explains.

And with data bulks of Skoltech this sphere is interesting to Maria Pukalchik from the Center for scientific and engineering computing technologies to tasks because machine learning and AI allow more precisely, more reliably and quicker to fulfill data and to receive more correct outputs. Besides, it is fashionable, she added.

See Also