“I think team ideas are the most important thing when developing a product”
Habr, hello! We continue a series of interviews with graduates of Newprolab, in which they tell about their history of transition to data science. The stories are different and will be interesting to those who are thinking about changing their career trajectory or how new knowledge can help solve current problems. Recently met with Yana Charuyskaya, Product Owner at MTS. Yana told how she came to the big data, how she grew professionally, remembered about her favorite project, which beside her knowledge and experience, also gave her friends. She told about the working atmosphere in MTS, about the projects that her team makes, about her dream, plans for the future, etc.
- Yana, tell us a little about yourself and your background.
- My name is Yana Charuyskaya, I am a Product Owner in MTS. I am interested in the area of Big Data and have been doing it for about two years. If in brief about my story: I graduated from HSE in the direction of “Business Informatics”, I studied for 6 years, then I studied for a year as a psychologist. For four years I worked in IT-consulting, three of them were engaged in data warehouses, data marts, building management reports mainly for large banks. The last year in consulting was engaged in machine learning and predictive analytics. Now I work at MTS as a product manager, I have a team of 6 people, and it’s growing, I’ll hire 7 more in the near future. In general, the company is also expanding, now there are more than 150 Big Data specialists in MTS and a lot of vacancies are open (we plan to increase state almost 2 times!). We develop several products with the team at the same time.
- Why and at what point did you decide to link your work with big data?
- At some point it became fashionable and interesting, but for me it was a difficult and incomprehensible area. Of course, I was at the university C # programming language and a little bit understood its theoretical foundations, but I never programmed myself. After getting a job in IT consulting, I wrote a lot of SQL scripts. But what is Python, neural networks, what is it to write programs in some programming language or to build predictive models - all this was for me a strange set of words that I really wanted to understand. It was a challenge for me, and I wanted to try. It all started when I found some Python sites on the Internet, started to train and solve simple puzzles. Decided, it seems to be somehow obtained, but something was missing. I found myself a Python tutor, with which we wrote code for solving an arbitrary system of linear equations by the Gauss method. I remember, we solved this problem about a month; I, however, did not work well, perhaps the teacher was not very good, I don’t know, but I finally decided it myself.
After an unsuccessful tutoring experience, I began to consider options for attending courses, I found the Big Data Specialist program on the Internet , and I was very glad that this was exactly what I needed: for three months to do machine learning and a great opportunity to get acquainted with a large number of systems for deploying big data. For me, online training is definitely not the most appropriate option, it is important for me not to sit at home alone at the computer, but to be in the company of people who are engaged in the same task, some element of competition must be necessary for you to do better than yours. colleague. Therefore, I chose Newprolab for myself and have no regrets about it.
At that time I was developing a data warehouse, for me it was already a little bit boring, I wanted to move into a new sphere, but the manager said that at that moment there was no such opportunity, however, he suggested that I completely close the analytics in a large bank. During the program, I realized that I still wanted to engage in machine learning, looked like an interview, looked for a job, I was given two offers. I come to the manager with them and say that I am leaving because I want to study Data Science. Then only he gave me this opportunity within the company. One of the conditions for moving to another area was the rejection of its team of analysts. I was left alone, it was hard. For the most part, I was engaged in pre-sales, that is, to make a model, you had to first find a client, sell this model, make it, protect and get paid for it. But these are some one-time activities, you can’t find a team for this, there was not very much expertise. Products were mostly commercial, we practically did not use open source solutions, so neither Python nor Spark were needed, models were built mainly with the help of commercial solutions for building classical response models. Due to the fact that I wanted to get more expertise in Data Science, create interesting products and work in a large team of specialists, I decided to look for a job again. the models were built mainly with the help of commercial solutions for the construction of classical response models. Due to the fact that I wanted to get more expertise in Data Science, create interesting products and work in a large team of specialists, I decided to look for a job again. the models were built mainly with the help of commercial solutions for the construction of classical response models. Due to the fact that I wanted to get more expertise in Data Science, create interesting products and work in a large team of specialists, I decided to look for a job again.
- We'll talk more than you lured to the MTS. Tell me what it is possible and necessary to retain good specialists, and is it worth it?
- Of course, it is worth it, and even better not to hold it, but to create all the conditions so that
they would like to stay! There are not many good specialists in the big data market, so I spend a lot of time maintaining a friendly atmosphere within the team. We communicate a lot, share ideas and impressions. We also go to conferences together and play intellectual games (for example, “What? Where? When?”). I try to give all the guys interesting tasks and watch them load so that there are no reworkings.
- And what difficulties did you face professionally at the very beginning, what challenges did you have to overcome?
- The biggest challenge was the programming language, because I am more a mathematician, and in programming there is another logic: the assignment of variables, the construction of classes, inheritance, polymorphism, and so on. The fact that programming is not mine, I decided for myself still in HSE. One of the biggest difficulties was to overcome the psychological barrier that I can also write code, and this is not a problem for me. In general, there were not many difficulties, there were many questions. It’s good that I had a lot of friends who answered all these questions: both my classmates in Newprolab and future friends I met at various conferences on Data Science and Big Data. And Open Data Science in Slack, where you can ask any question, and Data Science breakfasts, to which you can come and discuss any problem. In general, it seems to me
I communicate a lot with people, including novices in the field of Data Science, who doubt whether to go into the realm or not. Already all their life they work in some field, they are interested in Data Science, but they doubt whether it is worth changing something, they are afraid. I believe that if you want to change your life and go for your dream, it is quite real. I started with a promoter, worked in Auchan, advertised yogurts, then became a math tutor, worked for three years (and maybe more) in tutoring, but I understood that this brings some kind of income, but not all the time. I went to work at the leasing company as an economist, there was no IT, there was Excel at best, we didn't write macros either, the work was boring for me, and I was very worried that I was degrading. I tried to find myself in another area (actually, more associated with my education) - went to consulting, was engaged in storage. Then I got tired of the vaults, and again I had a choice as to where to go next. With such gradual steps related to changes in my professional activity, I came to Big Data, which I don’t regret. I was ready to waste my resources, my time, in order to understand this area. I think that if there is motivation, you can easily overcome all obstacles and achieve what you want. Once again, do not be afraid. I think that if there is motivation, you can easily overcome all obstacles and achieve what you want. Once again, do not be afraid. I think that if there is motivation, you can easily overcome all obstacles and achieve what you want. Once again, do not be afraid.
- Excellent life position and your story - a great example of the fact that if you want anything is possible. Returning to those who want to go to Data Science, what do you think, besides fear, what else can you stop? You communicate a lot with people, maybe they shared with you.
- The main thing - "I have no experience, I am not ready, I do not know anything." From my own experience I will immediately say: I went to the Newprolab courses, I studied there for two weeks and I already had two offers in the field of Data Science for good salaries. Two offers, and I was still studying! I didn’t even work in this area, I taught only a little bit of Python and just started to attend courses right now. I came to the employer and said that I am studying now on the program, on June 8 I will finish, I am motivated to develop in this area, I have relevant experience in data warehousing. Companies were ready to take me. Now the market is very narrow, there are very few data scientists, so companies usually take people to grow. If they see potential in you, they are ready to develop it.
After all, there are so many different training resources: Coursera , EdX , Udacity , to pump your knowledge. Even if you do not know statistics, you do not know linear algebra, mathematics, a programming language, you don’t know anything at all, for each of your ignorance there is a certain course that you can listen to quickly and quickly figure everything out, here the main desire and aspiration. And there is no such thing that “I have no experience,” the main thing is motivation, resources and energy. And I think there is time, if you want it.
According to Data Science, there are a lot of online courses now divorced, everywhere contextual advertising pops me up to one course or another. And their cost is rather big, and I see and hear the supplier of the courses for the first time. In general, this, of course, is HYIP, and I think that there are many low-quality courses that give practically nothing.
- From your observations: what soft and hard skills are often lacking for both beginners and experienced data scientists to become really high-class specialists? What should I look for?
- Very often there is a lack of practical skills for implementing models across the company, it is important to understand the subject area and properly prioritize the work. It is not worth spending a lot of time on a task, the results of solving which will not bring a positive effect on the company's activities. A data scientist is also advisable to develop their communication skills to present the results of their products both internally and externally. With regard to hard skills, I would like the candidates to better understand the terminology, understand the mathematical foundations of building models and know the cases of using models for various types of machine learning tasks. And creativity and imagination are also very important for developing new approaches to solving a problem (be it the addition of metrics to the data mart,
- Tell me more about the Data Science projects you have done.
- First, I will briefly tell you what was involved in consulting. We had projects in various fields, the department was not very large, and we were engaged in divers of different tasks. My first task was related to the model of response to a credit product in a large Russian bank. The model was successful, it gave a positive result, I did it with the help of a commercial decision; thanks to the implementation of this model, I was able to go through the full range of work on the harmonization of business requirements, the construction and production of the model, as well as assessing its quality and setting the rules. Since my past company specializes mainly in the banking sector, we mainly built models for banks, but also tried other areas (for example, insurance and retail). By the time I was not only as a data scientist in these projects, but also as a manager. It seems to me that the subject area can not be limited to, in any subject area, you can quickly understand. I am very glad that IT-consulting gave me such flexibility.
- Maybe there is some project or several projects that you are especially pleased to remember?
- Yes, there is such - my very first project in a large Russian bank, we had a very friendly team, we built a data warehouse from scratch, were engaged in its development, supported, built reports on it. It was a very cool product. A lot of experience gained, we have formed a great team. We have long been scattered across different companies, but we still actively maintain relationships. In this bank, we found ourselves, I guess.
- Good. Let's go to the MTS. Why exactly they? What is so interesting offered to do? What tasks do you and your team have now?
- First, I was attracted to the MTS by a huge team of Big Data, a bunch of specialists with whom you can consult at any time, which was not in IT consulting, but I was terribly lacking. We had a very experienced leader and several data scientists, it is clear that their experience was not enough to solve any problems. Roughly speaking, we had a standard set of tasks, which we did, and we tried not to deviate from this set of tasks, because we had no expertise. I am very glad that I chose MTS, we now have more than 150 people and we still want to grow by 70% by the end of the year. This is very cool, I love to communicate and share experiences, I think that new blood will definitely not hurt.
Secondly, there is a wide stack of technologies, we use Open Source: Python, Spark, Hive, Kafka - all popular words in the field of Big Data. We even have a commercial solution, but we don’t touch it and do not build models there. It's great that I managed to get acquainted with this stack on the Newprolab program and consolidate my knowledge later on at MTS.
Plus, of course, interesting tasks, interesting products. Customers are mostly domestic, but some of the products we bring out. Our team has several areas: strategic, it is tied to the implementation of models that are not necessarily currently bring us money; There are commercial projects that this year should show a financial result. I work in the R & D team, we sell products that in the future will help MTS to become better.
I and my team now have three products. The first is an assessment of the quality of service of our subscribers at various points of contact, including forecasting NPS (customer loyalty index - author's note) at the level of each subscriber. We have surveys that we conduct monthly for all our subscribers in order to understand whether they are ready to recommend the MTS brand or not. 0 - are not ready to recommend to anyone, 10 - are ready and actively do it. We collect these estimates and predict the assessment that the subscriber would give us if he had completed the survey, as well as we see the reasons that could affect this assessment; we can quickly help to fix them. This is the first product.
The second product is associated with voice analytics. Here, while only R & D, one of the tasks of voice analytics is speech-to-text recognition by calls to a contact center in order to analyze and automatically classify calls. At the moment, this is done by the operator, and the subject of messages may not always be sufficiently accurate.
About the third product, perhaps, I'll tell you later at some Big Data conference.
The team is very cool, we try to maintain a home working atmosphere, so that everyone is comfortable. I try to listen to each team member, everyone shares their ideas. I think team ideas are the most important when developing a product. In general, we also try to implement the most crazy ideas.
- Give an example of crazy ideas.
- It seems to me that our product by voice began this way. We did NPS, analyzed the estimates of our subscribers, and then someone asked: “Why can't we analyze voice calls to the call center?” Indeed, why not? We warn our subscribers that we can record and analyze. We ourselves do not listen to them, but thanks to the machining we can pull out the subject of calls to improve the quality of customer service.
It’s hard for me to give some concrete examples - any working moments when guys want to test something, try to implement something, or optimize it somewhere. We also try various solutions, many suppliers come to us, and offer the latest technologies. We carry out pilots with them, we look at the results.
- You, in addition to the MTS, considered some other options. What is crucial for you when choosing an employer?
- The openness of the company is important to me; I like the fact that I can consult with my colleagues, with my supervisor, share my fears, I know that he will understand and be able to give practical advice. Company reputation is important to me. I am ready, of course, to go to a startup if they have an interesting idea, but, in general, the company's reputation is important to me. I like working at MTS, we are the largest operator in Russia. I think that development opportunities are important too, and MTS encourages participation in various conferences both as a speaker and a listener. We have internal courses, which is very good, because I didn’t have enough of this at the previous place of work.
I need a flexible work schedule and a minimum of bureaucracy. In MTS, we practically have no bureaucracy, we do not write tons of paper, we have all the documentation being done through Confluence and Jira. We understand that at the production stage we will have to write some kind of TK, but, in general, we are fine with the documentation and with the coordination process too. I also like comfort, comfortable clothing is important for me, so that I don’t have to wear a jacket and tight shoes.
- At the Big Data Specialist you won first place at the end of the program, and at Deep Learning - the third for the project. Question: is it your assiduity / maybe a complex of excellent pupils, I don’t know if you have a serious approach to training / just an accident?
- Maybe it's all together. I had a goal, and I was going to it. She wanted to immerse herself in machine learning as soon as possible; Of course, combining the course with work was difficult, especially doing laboratory work and course projects, going to each lesson in time to write a test and get another small increase in points. Probably, I have a small complex of pupils, but if something does not work, I do not worry about it. When I do my work, I try to do it well, but if it doesn't work, I don’t get upset (at least I did my best). I think that serious intentions and a bit of luck also helped, because we had a very cool and responsive group, and also a smart coordinator on the program who answered all my questions and gave hints. Did not write directly what should have been done but correctly directed me in the right direction. I remember, we once even postponed the deadline, because we did not have time, and this helped us to pass the next laboratory test, which, we already thought, no one would pass (it was the most difficult).
And with “Deep Learning”, it’s probably just lucky. We only had a week course, on Friday there was the penultimate lesson, and on Saturday - the final one. And this Friday at the lecture I just decided to try one of the pre-trained nets, one of the newest ones. I looked at a comparative analysis of the trained nets in Keras, selected the best one and experimented with it a little. While everyone was learning, I launched Xception, which gave the maximum result, and made our two boys try to beat this result all night. There were some problems with the installation of this mesh, but they were quickly resolved, and the boys felt that it simply did not work on the Python in which we were trained. But I did it, and the guys tried to kill my result all night, so I'm in third place, not first.
- What are your thoughts and plans for the future? What do you want to study further? What skills and knowledge do you lack?
- I am now a Product Owner, and I want to further develop in this direction. First of all, I want to pump public speaking skills to a large audience. I feel that I worry when everyone is looking at me, and I forget the text, so I want to pump this skill in the first place. I spoke at the MTS conference in Sochi in August. This was my first performance in front of such a large audience. I wrote the text, taught it, went out to speak with pieces of paper, read something from there, told something, was very worried. I had a microphone in one hand, a presenter and papers in the other. As a result, the paper fell, the presentation accidentally turned off. It was a bit awkward, but everyone laughed with me.
Secondly, I plan to develop managerial skills in the field of people management. There is no doubt that there is always something to learn in this area.
In Newprolab I managed to complete three courses: “Data Engineer” , “Deep Learning”, “Big Data Specialist”. Including thanks to the experience of working as an analyst, I now know what my people are doing in the team; I can correctly set tasks; I understand how many tasks will be done. This is a big plus, I am sure that we will achieve good results.
- Traditionally, the question about the team, individuals in the industry of Big Data / Data Science, not only in Russia but also in the world. Are there people who inspire you, whose work do you follow?
- It's hard for me to say. It seems to me that I have a certain mindset and character, I am a more practical person: I have a task, I try to solve it. If I have questions about the implementation of my product, then I find articles on the Internet or at arxiv.organd share with the team the most interesting. I also go to Data Science for sharing experience, listening to conferences and courses, I read a lot of professional literature. There is no specific person whom I follow, or whose updates I constantly follow, but there are topics that interest me, I am ready to discuss them, I have many acquaintances in this area with whom I can discuss current problems and ask questions: both classmates at Newprolab, and people I met at conferences, and you can also talk at work, we have a large team of data scientists and data engineers.
- Do you have some kind of professional dream?
- I have been interested in psychology for about five years, I finished specialized courses, I studied in psychological groups and underwent individual therapy. I am interested in the area at the junction of psychology and machine learning. Perhaps in the future, when there is time, I will be able to create some product that will help people understand their emotions better and provide timely advice for making the right decision. In general, it would be interesting for me in the future to make a model that, on the basis of human psychology, would help people in everyday life. I do not know how realistic this is, but I would really like to do this.
- Well. You have already named one resource, maybe there are still some interesting profile blogs, telegram channels that you read at your leisure?
- I am sitting in chat Science Data in a telegram, but there are a lot of newbies and deep topics there. There is an ODS channel in the telegram, there you can read all the news, as in Slack. I am subscribed to some channels and news feeds that are being conducted by my friends - Grisha Sapunov, Kolya Markov, Petya Ermakov and a few others. And so not particularly watching the channels. If I need something, I'm looking at github, stackoverflow , at arxiv.org.
- Yana, tell us a little about yourself and your background.
- My name is Yana Charuyskaya, I am a Product Owner in MTS. I am interested in the area of Big Data and have been doing it for about two years. If in brief about my story: I graduated from HSE in the direction of “Business Informatics”, I studied for 6 years, then I studied for a year as a psychologist. For four years I worked in IT-consulting, three of them were engaged in data warehouses, data marts, building management reports mainly for large banks. The last year in consulting was engaged in machine learning and predictive analytics. Now I work at MTS as a product manager, I have a team of 6 people, and it’s growing, I’ll hire 7 more in the near future. In general, the company is also expanding, now there are more than 150 Big Data specialists in MTS and a lot of vacancies are open (we plan to increase state almost 2 times!). We develop several products with the team at the same time.
- Why and at what point did you decide to link your work with big data?
- At some point it became fashionable and interesting, but for me it was a difficult and incomprehensible area. Of course, I was at the university C # programming language and a little bit understood its theoretical foundations, but I never programmed myself. After getting a job in IT consulting, I wrote a lot of SQL scripts. But what is Python, neural networks, what is it to write programs in some programming language or to build predictive models - all this was for me a strange set of words that I really wanted to understand. It was a challenge for me, and I wanted to try. It all started when I found some Python sites on the Internet, started to train and solve simple puzzles. Decided, it seems to be somehow obtained, but something was missing. I found myself a Python tutor, with which we wrote code for solving an arbitrary system of linear equations by the Gauss method. I remember, we solved this problem about a month; I, however, did not work well, perhaps the teacher was not very good, I don’t know, but I finally decided it myself.
After an unsuccessful tutoring experience, I began to consider options for attending courses, I found the Big Data Specialist program on the Internet , and I was very glad that this was exactly what I needed: for three months to do machine learning and a great opportunity to get acquainted with a large number of systems for deploying big data. For me, online training is definitely not the most appropriate option, it is important for me not to sit at home alone at the computer, but to be in the company of people who are engaged in the same task, some element of competition must be necessary for you to do better than yours. colleague. Therefore, I chose Newprolab for myself and have no regrets about it.
At that time I was developing a data warehouse, for me it was already a little bit boring, I wanted to move into a new sphere, but the manager said that at that moment there was no such opportunity, however, he suggested that I completely close the analytics in a large bank. During the program, I realized that I still wanted to engage in machine learning, looked like an interview, looked for a job, I was given two offers. I come to the manager with them and say that I am leaving because I want to study Data Science. Then only he gave me this opportunity within the company. One of the conditions for moving to another area was the rejection of its team of analysts. I was left alone, it was hard. For the most part, I was engaged in pre-sales, that is, to make a model, you had to first find a client, sell this model, make it, protect and get paid for it. But these are some one-time activities, you can’t find a team for this, there was not very much expertise. Products were mostly commercial, we practically did not use open source solutions, so neither Python nor Spark were needed, models were built mainly with the help of commercial solutions for building classical response models. Due to the fact that I wanted to get more expertise in Data Science, create interesting products and work in a large team of specialists, I decided to look for a job again. the models were built mainly with the help of commercial solutions for the construction of classical response models. Due to the fact that I wanted to get more expertise in Data Science, create interesting products and work in a large team of specialists, I decided to look for a job again. the models were built mainly with the help of commercial solutions for the construction of classical response models. Due to the fact that I wanted to get more expertise in Data Science, create interesting products and work in a large team of specialists, I decided to look for a job again.
- We'll talk more than you lured to the MTS. Tell me what it is possible and necessary to retain good specialists, and is it worth it?
- Of course, it is worth it, and even better not to hold it, but to create all the conditions so that
they would like to stay! There are not many good specialists in the big data market, so I spend a lot of time maintaining a friendly atmosphere within the team. We communicate a lot, share ideas and impressions. We also go to conferences together and play intellectual games (for example, “What? Where? When?”). I try to give all the guys interesting tasks and watch them load so that there are no reworkings.
- And what difficulties did you face professionally at the very beginning, what challenges did you have to overcome?
- The biggest challenge was the programming language, because I am more a mathematician, and in programming there is another logic: the assignment of variables, the construction of classes, inheritance, polymorphism, and so on. The fact that programming is not mine, I decided for myself still in HSE. One of the biggest difficulties was to overcome the psychological barrier that I can also write code, and this is not a problem for me. In general, there were not many difficulties, there were many questions. It’s good that I had a lot of friends who answered all these questions: both my classmates in Newprolab and future friends I met at various conferences on Data Science and Big Data. And Open Data Science in Slack, where you can ask any question, and Data Science breakfasts, to which you can come and discuss any problem. In general, it seems to me
I communicate a lot with people, including novices in the field of Data Science, who doubt whether to go into the realm or not. Already all their life they work in some field, they are interested in Data Science, but they doubt whether it is worth changing something, they are afraid. I believe that if you want to change your life and go for your dream, it is quite real. I started with a promoter, worked in Auchan, advertised yogurts, then became a math tutor, worked for three years (and maybe more) in tutoring, but I understood that this brings some kind of income, but not all the time. I went to work at the leasing company as an economist, there was no IT, there was Excel at best, we didn't write macros either, the work was boring for me, and I was very worried that I was degrading. I tried to find myself in another area (actually, more associated with my education) - went to consulting, was engaged in storage. Then I got tired of the vaults, and again I had a choice as to where to go next. With such gradual steps related to changes in my professional activity, I came to Big Data, which I don’t regret. I was ready to waste my resources, my time, in order to understand this area. I think that if there is motivation, you can easily overcome all obstacles and achieve what you want. Once again, do not be afraid. I think that if there is motivation, you can easily overcome all obstacles and achieve what you want. Once again, do not be afraid. I think that if there is motivation, you can easily overcome all obstacles and achieve what you want. Once again, do not be afraid.
- Excellent life position and your story - a great example of the fact that if you want anything is possible. Returning to those who want to go to Data Science, what do you think, besides fear, what else can you stop? You communicate a lot with people, maybe they shared with you.
- The main thing - "I have no experience, I am not ready, I do not know anything." From my own experience I will immediately say: I went to the Newprolab courses, I studied there for two weeks and I already had two offers in the field of Data Science for good salaries. Two offers, and I was still studying! I didn’t even work in this area, I taught only a little bit of Python and just started to attend courses right now. I came to the employer and said that I am studying now on the program, on June 8 I will finish, I am motivated to develop in this area, I have relevant experience in data warehousing. Companies were ready to take me. Now the market is very narrow, there are very few data scientists, so companies usually take people to grow. If they see potential in you, they are ready to develop it.
After all, there are so many different training resources: Coursera , EdX , Udacity , to pump your knowledge. Even if you do not know statistics, you do not know linear algebra, mathematics, a programming language, you don’t know anything at all, for each of your ignorance there is a certain course that you can listen to quickly and quickly figure everything out, here the main desire and aspiration. And there is no such thing that “I have no experience,” the main thing is motivation, resources and energy. And I think there is time, if you want it.
According to Data Science, there are a lot of online courses now divorced, everywhere contextual advertising pops me up to one course or another. And their cost is rather big, and I see and hear the supplier of the courses for the first time. In general, this, of course, is HYIP, and I think that there are many low-quality courses that give practically nothing.
- From your observations: what soft and hard skills are often lacking for both beginners and experienced data scientists to become really high-class specialists? What should I look for?
- Very often there is a lack of practical skills for implementing models across the company, it is important to understand the subject area and properly prioritize the work. It is not worth spending a lot of time on a task, the results of solving which will not bring a positive effect on the company's activities. A data scientist is also advisable to develop their communication skills to present the results of their products both internally and externally. With regard to hard skills, I would like the candidates to better understand the terminology, understand the mathematical foundations of building models and know the cases of using models for various types of machine learning tasks. And creativity and imagination are also very important for developing new approaches to solving a problem (be it the addition of metrics to the data mart,
- Tell me more about the Data Science projects you have done.
- First, I will briefly tell you what was involved in consulting. We had projects in various fields, the department was not very large, and we were engaged in divers of different tasks. My first task was related to the model of response to a credit product in a large Russian bank. The model was successful, it gave a positive result, I did it with the help of a commercial decision; thanks to the implementation of this model, I was able to go through the full range of work on the harmonization of business requirements, the construction and production of the model, as well as assessing its quality and setting the rules. Since my past company specializes mainly in the banking sector, we mainly built models for banks, but also tried other areas (for example, insurance and retail). By the time I was not only as a data scientist in these projects, but also as a manager. It seems to me that the subject area can not be limited to, in any subject area, you can quickly understand. I am very glad that IT-consulting gave me such flexibility.
- Maybe there is some project or several projects that you are especially pleased to remember?
- Yes, there is such - my very first project in a large Russian bank, we had a very friendly team, we built a data warehouse from scratch, were engaged in its development, supported, built reports on it. It was a very cool product. A lot of experience gained, we have formed a great team. We have long been scattered across different companies, but we still actively maintain relationships. In this bank, we found ourselves, I guess.
- Good. Let's go to the MTS. Why exactly they? What is so interesting offered to do? What tasks do you and your team have now?
- First, I was attracted to the MTS by a huge team of Big Data, a bunch of specialists with whom you can consult at any time, which was not in IT consulting, but I was terribly lacking. We had a very experienced leader and several data scientists, it is clear that their experience was not enough to solve any problems. Roughly speaking, we had a standard set of tasks, which we did, and we tried not to deviate from this set of tasks, because we had no expertise. I am very glad that I chose MTS, we now have more than 150 people and we still want to grow by 70% by the end of the year. This is very cool, I love to communicate and share experiences, I think that new blood will definitely not hurt.
Secondly, there is a wide stack of technologies, we use Open Source: Python, Spark, Hive, Kafka - all popular words in the field of Big Data. We even have a commercial solution, but we don’t touch it and do not build models there. It's great that I managed to get acquainted with this stack on the Newprolab program and consolidate my knowledge later on at MTS.
Plus, of course, interesting tasks, interesting products. Customers are mostly domestic, but some of the products we bring out. Our team has several areas: strategic, it is tied to the implementation of models that are not necessarily currently bring us money; There are commercial projects that this year should show a financial result. I work in the R & D team, we sell products that in the future will help MTS to become better.
I and my team now have three products. The first is an assessment of the quality of service of our subscribers at various points of contact, including forecasting NPS (customer loyalty index - author's note) at the level of each subscriber. We have surveys that we conduct monthly for all our subscribers in order to understand whether they are ready to recommend the MTS brand or not. 0 - are not ready to recommend to anyone, 10 - are ready and actively do it. We collect these estimates and predict the assessment that the subscriber would give us if he had completed the survey, as well as we see the reasons that could affect this assessment; we can quickly help to fix them. This is the first product.
The second product is associated with voice analytics. Here, while only R & D, one of the tasks of voice analytics is speech-to-text recognition by calls to a contact center in order to analyze and automatically classify calls. At the moment, this is done by the operator, and the subject of messages may not always be sufficiently accurate.
About the third product, perhaps, I'll tell you later at some Big Data conference.
The team is very cool, we try to maintain a home working atmosphere, so that everyone is comfortable. I try to listen to each team member, everyone shares their ideas. I think team ideas are the most important when developing a product. In general, we also try to implement the most crazy ideas.
- Give an example of crazy ideas.
- It seems to me that our product by voice began this way. We did NPS, analyzed the estimates of our subscribers, and then someone asked: “Why can't we analyze voice calls to the call center?” Indeed, why not? We warn our subscribers that we can record and analyze. We ourselves do not listen to them, but thanks to the machining we can pull out the subject of calls to improve the quality of customer service.
It’s hard for me to give some concrete examples - any working moments when guys want to test something, try to implement something, or optimize it somewhere. We also try various solutions, many suppliers come to us, and offer the latest technologies. We carry out pilots with them, we look at the results.
- You, in addition to the MTS, considered some other options. What is crucial for you when choosing an employer?
- The openness of the company is important to me; I like the fact that I can consult with my colleagues, with my supervisor, share my fears, I know that he will understand and be able to give practical advice. Company reputation is important to me. I am ready, of course, to go to a startup if they have an interesting idea, but, in general, the company's reputation is important to me. I like working at MTS, we are the largest operator in Russia. I think that development opportunities are important too, and MTS encourages participation in various conferences both as a speaker and a listener. We have internal courses, which is very good, because I didn’t have enough of this at the previous place of work.
I need a flexible work schedule and a minimum of bureaucracy. In MTS, we practically have no bureaucracy, we do not write tons of paper, we have all the documentation being done through Confluence and Jira. We understand that at the production stage we will have to write some kind of TK, but, in general, we are fine with the documentation and with the coordination process too. I also like comfort, comfortable clothing is important for me, so that I don’t have to wear a jacket and tight shoes.
- At the Big Data Specialist you won first place at the end of the program, and at Deep Learning - the third for the project. Question: is it your assiduity / maybe a complex of excellent pupils, I don’t know if you have a serious approach to training / just an accident?
- Maybe it's all together. I had a goal, and I was going to it. She wanted to immerse herself in machine learning as soon as possible; Of course, combining the course with work was difficult, especially doing laboratory work and course projects, going to each lesson in time to write a test and get another small increase in points. Probably, I have a small complex of pupils, but if something does not work, I do not worry about it. When I do my work, I try to do it well, but if it doesn't work, I don’t get upset (at least I did my best). I think that serious intentions and a bit of luck also helped, because we had a very cool and responsive group, and also a smart coordinator on the program who answered all my questions and gave hints. Did not write directly what should have been done but correctly directed me in the right direction. I remember, we once even postponed the deadline, because we did not have time, and this helped us to pass the next laboratory test, which, we already thought, no one would pass (it was the most difficult).
And with “Deep Learning”, it’s probably just lucky. We only had a week course, on Friday there was the penultimate lesson, and on Saturday - the final one. And this Friday at the lecture I just decided to try one of the pre-trained nets, one of the newest ones. I looked at a comparative analysis of the trained nets in Keras, selected the best one and experimented with it a little. While everyone was learning, I launched Xception, which gave the maximum result, and made our two boys try to beat this result all night. There were some problems with the installation of this mesh, but they were quickly resolved, and the boys felt that it simply did not work on the Python in which we were trained. But I did it, and the guys tried to kill my result all night, so I'm in third place, not first.
- What are your thoughts and plans for the future? What do you want to study further? What skills and knowledge do you lack?
- I am now a Product Owner, and I want to further develop in this direction. First of all, I want to pump public speaking skills to a large audience. I feel that I worry when everyone is looking at me, and I forget the text, so I want to pump this skill in the first place. I spoke at the MTS conference in Sochi in August. This was my first performance in front of such a large audience. I wrote the text, taught it, went out to speak with pieces of paper, read something from there, told something, was very worried. I had a microphone in one hand, a presenter and papers in the other. As a result, the paper fell, the presentation accidentally turned off. It was a bit awkward, but everyone laughed with me.
Secondly, I plan to develop managerial skills in the field of people management. There is no doubt that there is always something to learn in this area.
In Newprolab I managed to complete three courses: “Data Engineer” , “Deep Learning”, “Big Data Specialist”. Including thanks to the experience of working as an analyst, I now know what my people are doing in the team; I can correctly set tasks; I understand how many tasks will be done. This is a big plus, I am sure that we will achieve good results.
- Traditionally, the question about the team, individuals in the industry of Big Data / Data Science, not only in Russia but also in the world. Are there people who inspire you, whose work do you follow?
- It's hard for me to say. It seems to me that I have a certain mindset and character, I am a more practical person: I have a task, I try to solve it. If I have questions about the implementation of my product, then I find articles on the Internet or at arxiv.organd share with the team the most interesting. I also go to Data Science for sharing experience, listening to conferences and courses, I read a lot of professional literature. There is no specific person whom I follow, or whose updates I constantly follow, but there are topics that interest me, I am ready to discuss them, I have many acquaintances in this area with whom I can discuss current problems and ask questions: both classmates at Newprolab, and people I met at conferences, and you can also talk at work, we have a large team of data scientists and data engineers.
- Do you have some kind of professional dream?
- I have been interested in psychology for about five years, I finished specialized courses, I studied in psychological groups and underwent individual therapy. I am interested in the area at the junction of psychology and machine learning. Perhaps in the future, when there is time, I will be able to create some product that will help people understand their emotions better and provide timely advice for making the right decision. In general, it would be interesting for me in the future to make a model that, on the basis of human psychology, would help people in everyday life. I do not know how realistic this is, but I would really like to do this.
- Well. You have already named one resource, maybe there are still some interesting profile blogs, telegram channels that you read at your leisure?
- I am sitting in chat Science Data in a telegram, but there are a lot of newbies and deep topics there. There is an ODS channel in the telegram, there you can read all the news, as in Slack. I am subscribed to some channels and news feeds that are being conducted by my friends - Grisha Sapunov, Kolya Markov, Petya Ermakov and a few others. And so not particularly watching the channels. If I need something, I'm looking at github, stackoverflow , at arxiv.org.