About Tinkoff.ru magistracy at MIPT
Hello, Habr! My name is Sasha Minochkina. I am finishing my graduate studies - I decided to tell how this is generally arranged. My article will be useful to young professionals who want to study and look towards the magistracy from Tinkoff.ru, but are afraid that they will be forced to push Oleg Tinkov at the office. I’ll tell you how I acted, studied and worked so that there were no such fears.
I studied at the Moscow Institute of Physics and Technology, studied bioinformatics. And then I wanted to know what machine learning was, and it started.
It all started with a course at MIPT. Then there was Fintech School 2017 at Tinkoff.ru. I definitely understood on it that I like all this and want to work here. In addition, in this company I had a lot of cool acquaintances - programmers. And this, as you know, is the best advertisement.
At the end of Fintech School, news appeared that Tinkoff.ru was opening a department of financial technology at Fiztekh. In May there was a presentation of the new department ( slidesfrom last year’s lecture). It was attended by representatives of the department and managers of various projects with whom it was possible to communicate. I learned that when enrolling in a department, they automatically get a job. The idea of combining study and work in one place was attractive.
I applied and waited for someone to be invited for an interview. Applications were accepted from May to early June. All interviews were in early June.
And so I was invited for an interview. It consisted of a conversation with a bank employee. Checked what knowledge is in linal, statistics, algorithms. They also looked at what I know about classic ML, neural networks and infrastructure. As in any interview, they were still interested in experience. The selection results were sent after a couple of weeks.
In total, 20 people are recruiting for the department. Many are interested in the question of how much you need to be cool to qualify. Of course, you need to have a good understanding of basic university mathematics and have an idea of what machine learning / functional programming is. Last year, many applicants who successfully passed the tests already had experience in implementing core projects.
All important dates can be found in the VK group and Telegram channel .
In addition to an interview at the department, it was necessary to enter the master's program at MIPT itself . Since I studied at the bachelor's program at Fiztekh, there were no problems with admission. But those who did not study had to work hard and pass exams in mathematics and computer science. According to applicants from other universities, video lectures for the preparation of different years are very helpful, and the tasks in the exam are identical to the tasks in video lectures.
Even in the magistracy, you can count publications in scientific journals, participation in conferences, prizes in the "I-Professional" Olympiad and other things. More information is here .
The department of Tinkoff.ru is based at the Moscow Institute of Physics and Technology, at the Physics and Technology School of Applied Mathematics and Computer Science (FPMI). It has 2 directions: applied mathematics and physics (PMF) and applied mathematics and computer science (PMI). There are no restrictions on the number of students in a particular field. The main thing is that in the amount of our students with PIP and PFM there should be 20 people.
The department has two specializations: machine learning and functional programming. There is no strict distribution of students, it all depends on the set. But in each of the flows of students in the first specialization, twice as much was obtained than in the second. By the way, I went to machine learning, PMI.
Couples in the magistracy are divided into two types: at the PhysTech and at the department.
Couples at the PhysTech.
Courses in the Master of Physics and Mathematics are very dependent on the direction. Strange as it may seem, most of the courses at the PMF are about finance and innovation: the mathematical theory of finance, evaluating the effectiveness of investment projects, the national innovation system, etc. At the PMI, courses are taught with a bias in advanced mathematics and data analysis: robust methods in statistics, ext. chapters of discrete mathematics, NLP, data visualization methods, etc.
As it turned out, the famous “PhysTech system” extends not only to undergraduate studies. And now I'm not talking about the optionalness of attending lectures, but about the fact that you can almost always replace courses from your curriculum with others that you like more. Only for graduate school it is important that the desired courses are also from the graduate school - couples from undergraduate classes will not work. For example, I exchanged several PMI items for PMF items, because for me they were more useful.
In the first and second semesters, Thursday and Saturday are given for couples, in the third semesters - only Thursday. All Thursday pairs are held in 1C building on Timiryazevskaya, and all Saturday pairs - at Fiztekh, in Dolgoprudny.
Couples at the Department
Most of the courses are faculty, not faculty. Cathedral pairs are more focused on the chosen specialization than faculty ones.
The main pairs:
I’ll tell you more about the subjects of my direction - “machine learning” and “deep learning”. Each course consists of lectures and seminars. It explains everything from the very basics to state-of-the-art algorithms. A lot of deep theory with statistics. But also a lot of practice, so this whole theory from lectures became understandable. In theory, you can come with zero knowledge. Then it will be necessary to put a lot of time and effort to care. But it's worth it :)
On "machine learning" tasks were solved using Scikit-learn. Homework was after each lecture, among them several Kaggle Inclass competitions. A couple of times we stayed in pairs until 23 pm, because the lecturer was very interested in the topic and tried to tell as much as possible.
In “deep learning” all tasks were solved on PyTorch. There was a main teacher and several invitees who were specialists in certain topics. It was the most voluminous course for the entire magistracy. And although many of us at the beginning of the course knew something about this topic, it was somehow quite difficult. Lectures and seminars on "deep learning" were recorded. It helped out more than once: both in preparation for exams, and during the semester.
Couples help a lot in work. Firstly, because we were given knowledge just in the specialization in which we work. Machine and deep learning courses help build working models. A Big Data course helps you do this effectively. A course on software architecture helps synchronize development with colleagues and raise services for the models we have written. Secondly, at the lectures we could ask any question about work and get advice: which metrics are better to use for a specific task, which models are worth trying, why nothing works at all.
All classes are held in the office, where we work, at the Water Stadium. No need to waste time on the road between work and school. And it can not but rejoice! Pairs are read in the evenings, after work: 2-3 hours from 18:00 to 21:00. Such a schedule allows you to devote more time to work. But sometimes, of course, it is difficult to perceive the material after a full day.
In general, if I have any questions about the educational process, I can write to the curators of the department 24/7 and they will help solve all the problems.
As I have already said, upon admission to the department they are hired. You can work from 24 hours a week. Most of my classmates at the beginning of my studies got exactly that much. But there were those who immediately went to full-time. It all depends on how much time you want to devote to study, and how much work. I started from 24 hours, after the first semester I switched to 32 hours and only after the second semester I switched to a full week.
After passing to the department in accordance with your skills and interests you are assigned to one of the teams.
Projects in the field of "machine learning":
Projects in the field of "functional programming":
Each student is given a mentor. This is usually the leader of the team you have chosen. He will help to get used to the company, propose tasks and contribute to your development. In principle, you can come with zero experience if you pass the selection to the department, but it is certainly important to quickly understand what you are doing.
Over time, you can move on to other projects, in other teams. For example, one of my classmates was engaged in the task of predicting cash at ATMs, and after completing this project he began to engage in chat bots. Here, no one restricts you in choosing.
Work is a good practice of what we are taught in the magistracy. I like the fact that the department does not graduate from specialists "divorced from reality" who know a lot. And those specialists who are also able to apply this knowledge.
Well, I almost forgot to talk about the diploma. We write a diploma on those projects that we do at work. It is very comfortable.
Tinkoff.ru Master's program is ideal for those who want to study and work a lot. And so that study and work organically complement each other. Yes, it happens! Tonight you are listening to a couple about the Advantage Actor Critic algorithm, tomorrow you will implement such a model at work. Couples will not interfere with work, because they are held in the same place as work in the evening.
If you want to sign up for an interview at the department, fill out the form .
“What if I'm still small for graduate school?”, “Already in graduate school” or something else. We have cool fintech schools twice a year, year-round internships and a laboratory at Fiztekh. Read more here . The latest news can be learned from the group in VK and the Telegram channel.
If you still have any questions, write, I will be happy to answer :)
Receipt. Why is Tinkoff.ru generally?
I studied at the Moscow Institute of Physics and Technology, studied bioinformatics. And then I wanted to know what machine learning was, and it started.
It all started with a course at MIPT. Then there was Fintech School 2017 at Tinkoff.ru. I definitely understood on it that I like all this and want to work here. In addition, in this company I had a lot of cool acquaintances - programmers. And this, as you know, is the best advertisement.
At the end of Fintech School, news appeared that Tinkoff.ru was opening a department of financial technology at Fiztekh. In May there was a presentation of the new department ( slidesfrom last year’s lecture). It was attended by representatives of the department and managers of various projects with whom it was possible to communicate. I learned that when enrolling in a department, they automatically get a job. The idea of combining study and work in one place was attractive.
I applied and waited for someone to be invited for an interview. Applications were accepted from May to early June. All interviews were in early June.
And so I was invited for an interview. It consisted of a conversation with a bank employee. Checked what knowledge is in linal, statistics, algorithms. They also looked at what I know about classic ML, neural networks and infrastructure. As in any interview, they were still interested in experience. The selection results were sent after a couple of weeks.
In total, 20 people are recruiting for the department. Many are interested in the question of how much you need to be cool to qualify. Of course, you need to have a good understanding of basic university mathematics and have an idea of what machine learning / functional programming is. Last year, many applicants who successfully passed the tests already had experience in implementing core projects.
All important dates can be found in the VK group and Telegram channel .
Admission to MIPT
In addition to an interview at the department, it was necessary to enter the master's program at MIPT itself . Since I studied at the bachelor's program at Fiztekh, there were no problems with admission. But those who did not study had to work hard and pass exams in mathematics and computer science. According to applicants from other universities, video lectures for the preparation of different years are very helpful, and the tasks in the exam are identical to the tasks in video lectures.
Even in the magistracy, you can count publications in scientific journals, participation in conferences, prizes in the "I-Professional" Olympiad and other things. More information is here .
Department structure
The department of Tinkoff.ru is based at the Moscow Institute of Physics and Technology, at the Physics and Technology School of Applied Mathematics and Computer Science (FPMI). It has 2 directions: applied mathematics and physics (PMF) and applied mathematics and computer science (PMI). There are no restrictions on the number of students in a particular field. The main thing is that in the amount of our students with PIP and PFM there should be 20 people.
The department has two specializations: machine learning and functional programming. There is no strict distribution of students, it all depends on the set. But in each of the flows of students in the first specialization, twice as much was obtained than in the second. By the way, I went to machine learning, PMI.
Couples
Couples in the magistracy are divided into two types: at the PhysTech and at the department.
Couples at the PhysTech.
Courses in the Master of Physics and Mathematics are very dependent on the direction. Strange as it may seem, most of the courses at the PMF are about finance and innovation: the mathematical theory of finance, evaluating the effectiveness of investment projects, the national innovation system, etc. At the PMI, courses are taught with a bias in advanced mathematics and data analysis: robust methods in statistics, ext. chapters of discrete mathematics, NLP, data visualization methods, etc.
As it turned out, the famous “PhysTech system” extends not only to undergraduate studies. And now I'm not talking about the optionalness of attending lectures, but about the fact that you can almost always replace courses from your curriculum with others that you like more. Only for graduate school it is important that the desired courses are also from the graduate school - couples from undergraduate classes will not work. For example, I exchanged several PMI items for PMF items, because for me they were more useful.
In the first and second semesters, Thursday and Saturday are given for couples, in the third semesters - only Thursday. All Thursday pairs are held in 1C building on Timiryazevskaya, and all Saturday pairs - at Fiztekh, in Dolgoprudny.
Couples at the Department
Most of the courses are faculty, not faculty. Cathedral pairs are more focused on the chosen specialization than faculty ones.
The main pairs:
- algorithms and data structures
- software architecture
- big data
- machine learning (machine learning direction)
- deep learning (machine learning direction)
- Scala (the direction of "functional programming")
I’ll tell you more about the subjects of my direction - “machine learning” and “deep learning”. Each course consists of lectures and seminars. It explains everything from the very basics to state-of-the-art algorithms. A lot of deep theory with statistics. But also a lot of practice, so this whole theory from lectures became understandable. In theory, you can come with zero knowledge. Then it will be necessary to put a lot of time and effort to care. But it's worth it :)
On "machine learning" tasks were solved using Scikit-learn. Homework was after each lecture, among them several Kaggle Inclass competitions. A couple of times we stayed in pairs until 23 pm, because the lecturer was very interested in the topic and tried to tell as much as possible.
In “deep learning” all tasks were solved on PyTorch. There was a main teacher and several invitees who were specialists in certain topics. It was the most voluminous course for the entire magistracy. And although many of us at the beginning of the course knew something about this topic, it was somehow quite difficult. Lectures and seminars on "deep learning" were recorded. It helped out more than once: both in preparation for exams, and during the semester.
Couples help a lot in work. Firstly, because we were given knowledge just in the specialization in which we work. Machine and deep learning courses help build working models. A Big Data course helps you do this effectively. A course on software architecture helps synchronize development with colleagues and raise services for the models we have written. Secondly, at the lectures we could ask any question about work and get advice: which metrics are better to use for a specific task, which models are worth trying, why nothing works at all.
All classes are held in the office, where we work, at the Water Stadium. No need to waste time on the road between work and school. And it can not but rejoice! Pairs are read in the evenings, after work: 2-3 hours from 18:00 to 21:00. Such a schedule allows you to devote more time to work. But sometimes, of course, it is difficult to perceive the material after a full day.
In general, if I have any questions about the educational process, I can write to the curators of the department 24/7 and they will help solve all the problems.
Work
As I have already said, upon admission to the department they are hired. You can work from 24 hours a week. Most of my classmates at the beginning of my studies got exactly that much. But there were those who immediately went to full-time. It all depends on how much time you want to devote to study, and how much work. I started from 24 hours, after the first semester I switched to 32 hours and only after the second semester I switched to a full week.
After passing to the department in accordance with your skills and interests you are assigned to one of the teams.
Projects in the field of "machine learning":
- speech recognition
- Nlp
- dialogue systems
- speech synthesis
- computer vision
- recommendation systems
- antifraud
- credit scoring
- recruitment automation
- predictive analytics
Projects in the field of "functional programming":
- Internet Bank of individuals
- Legal entities Internet Bank
- Trading platform
- Data management platform
- Identity graph
Each student is given a mentor. This is usually the leader of the team you have chosen. He will help to get used to the company, propose tasks and contribute to your development. In principle, you can come with zero experience if you pass the selection to the department, but it is certainly important to quickly understand what you are doing.
Over time, you can move on to other projects, in other teams. For example, one of my classmates was engaged in the task of predicting cash at ATMs, and after completing this project he began to engage in chat bots. Here, no one restricts you in choosing.
Work is a good practice of what we are taught in the magistracy. I like the fact that the department does not graduate from specialists "divorced from reality" who know a lot. And those specialists who are also able to apply this knowledge.
Well, I almost forgot to talk about the diploma. We write a diploma on those projects that we do at work. It is very comfortable.
Conclusion
Tinkoff.ru Master's program is ideal for those who want to study and work a lot. And so that study and work organically complement each other. Yes, it happens! Tonight you are listening to a couple about the Advantage Actor Critic algorithm, tomorrow you will implement such a model at work. Couples will not interfere with work, because they are held in the same place as work in the evening.
If you want to sign up for an interview at the department, fill out the form .
“What if I'm still small for graduate school?”, “Already in graduate school” or something else. We have cool fintech schools twice a year, year-round internships and a laboratory at Fiztekh. Read more here . The latest news can be learned from the group in VK and the Telegram channel.
If you still have any questions, write, I will be happy to answer :)