Sberseasons: how I spent this summer
Hello! What to do if you have not made plans for the summer? Internship! Sberbank offered me a two-month paid internship. In this text I will talk about the Sberseasons internship program itself, how I was selected, what I did and what I learned. This article may be useful for students completing math or economics studies.
What is Sberseasons?
Sberseasons is a paid internship conducted by Sberbank of 3 and 4 year undergraduate, postgraduate and postgraduate students three times a year.In fact, Sberbank offers to try one of the proposed areas (with the possibility of further employment) for two months, while receiving a salary. More information can be found here .
How was the selection
In the spring I finished the third year of the specialty “Fundamental Mathematics” and learned about a paid internship that Sberbank conducts. I saw the program "Java-development", which takes place in Moscow.
While reading various information, I saw that this internship was taking place not only in Moscow, but also in other Russian cities: St. Petersburg, Nizhny Novgorod, Yekaterinburg, Samara and Voronezh. When I saw my city (Samara), I immediately rejoiced that if I could manage to qualify, I would not have to go anywhere. But immediately after the joy came disappointment: in Samara there is no direction associated with Java. There were only economic models and analytics. After a bit of weighing all the pros and cons, I decided to try myself in analytics.
It all started with the fact that on May 12th I filled out a questionnaire (some information about myself) and my status went into the standby feedback mode. About a week later, I received an answer saying that I was invited for testing, scheduled for June 8th. The test consisted of two parts: in the first one, there were puzzles for easy math and logic (for example, finding the angle formed by the hands of the clock at a certain time, or figuring out how the employee’s salary would change if it went down percent and increase after that on percent); The second part included tasks on knowledge of the basic structures and functions of Excel and knowledge of some operators in the SQL language. After this testing, everyone was asked to wait: those who passed this stage would be invited to an in-person interview.
Almost two weeks later, on June 22, they called me and invited me to this most in-person interview at the Center for Support of Client Operations. I came, and, as it turned out, the interview itself will not. The head of the personnel department conducted an oral tour of the different departments of their center, told them who was doing what and asked where I would most like to go. I chose the IT department. We agreed that she would ask in this department if they needed an employee, after which she would call back and report the result.
The result was not long in coming. On the 29th, they finally invited me to the final interview with the head of the bank card operations department. I asked him that I could, that I didn’t (at the end it turned out that I couldn’t do anything that could be useful to them). He himself told about his department and about what they are doing. And he decided that he was ready to hire me for these two months. A week later, on July 6, I filled out documents for employment. The 9th was the first working day.
What i did
On Monday, they gave me a job and appointed a manager. After a little chat with me, he decided that we would do machine learning based on the Python code. Since I first encountered machine learning, the maximum I could write in Python is the solution to the A + B problem, we started with an introduction to various supporting Python libraries, with some algorithms and techniques for machine learning.
The first week I analyzed the notorious data about the passengers of the Titanic. This task was given so that I “played around” with the data, looked at the syntax of the code, at the methods that exist and facilitate the analysis, and tried myself in drawing the data output. After that, a model was drawn up which, according to the passenger information, concludes: most likely such a person would have survived in that situation or not. In other words, I did what was described in great detail here . Next week, I was introduced to other techniques that allow me to work with text. And with the help of them I analyzed the text of the book “War and Peace”. I figured out which word is more common and made a visual chart.
In parallel with the study of Python and machine learning, there were assignments from other members of the department. Basically, these tasks were simple: send emails to colleagues, make a comparison of columns in Excel using the CDF function, and so on; that is, I served as an assistant.
However, exactly one month after the beginning of the internship, I was taught another important event - the creation of requests. It is, in principle, easy, but very responsible. The point is that from time to time there is a need to change any information stored in the database: the department receives a request that describes the problem, the reasons for the change and what specifically needs to be changed. After that, someone from the staff (including me) created the Request for Change (ZNI), which included a description of the problem (including the reasons and what needs to be changed) and the script in SQL. After that, an internal approval for a change was conducted, this script was launched, executed, the data changed, and in fact the request was considered completed (closed).
Let's return to machine learning. Each such request has a topic and description. Further, these requests are distributed within the department by employee, depending on the topic. It is clear that the topic will directly depend on the description. Therefore, it was decided to create a model that will distribute requests by employees. When I came, such a model was already created, but it was unstable. I, together with the manager, “optimized” the model, that is, I selected more appropriate classifiers and parameters for them, worked on the input data, looked for errors, and so on.
The result was a model that defined the topic with precision%
What i learned
First of all, I got acquainted with machine learning, its algorithms and methods, learning not only on theory, but also trying it in practice (and on a real-life task).
The next thing I learned was SQL queries. Of course, I had heard of them before. But there was no need to start studying, but it turned out to be both interesting and useful.
The internal structure of the bank - another useful information in the piggy bank of life experience. During the internship some excursions to other buildings of Sberbank were organized, which described how Sberbank was arranged, its brief history and the closest goals. Meetings were also held aimed at successful adaptation of new employees, which took place in the form of a game or dialogue.
Finally, I met interesting and benevolent people who are ready to help you with any problem and at any time (one of my colleagues even offered to bring my own screwdriver from home so that I could tighten the bolts on the laptop, and the other on my day off helped me figure out ).
Impressions of the past
Overall, I liked everything. I gained knowledge that, in my opinion, are not useless (and even vice versa, are used in developing areas), I tried myself in one of the largest banks in the country, I became acquainted with its structure and interaction process — employee-employee and employee-client, made friends with new interesting people.