According to research by HeadHunter and Mail.ru, the demand for data scientists exceeds the supply, but even so, young specialists do not always manage to find a job. We tell you what the graduates of the courses lack and where to study for those who are planning a big career in Data Science.
βThey come and think that now they will earn 500k per second, because they know the names of the frameworks and how to run a two-line model from themβ
Emil Maharramov leads a group of computational chemistry services at biocad and at interviews is faced with the fact that candidates do not have a systematic understanding of the profession. They complete courses, come with well-pumped Python and SQL, can raise Hadoop or Spark in 2 seconds, complete the task according to a clear TOR. But at the same time, a step to the side is no longer there. Although it is the flexibility of solutions that employers expect from their specialists in the field of Data Science.
What is happening in the Data Science market
The competencies of young professionals reflect the situation in the labor market. Here, demand significantly exceeds supply, so desperate employers are often really ready to hire completely green specialists and grow them for themselves. The option is working, but it is only suitable if the team already has an experienced team leader who will take over the training of the junior.
According to a study by HeadHunter and Mail.ru, data scientists are among the most in demand on the market:
- In 2019, there were 9,6 times more vacancies in the field of data analysis, and 7,2 times more in the field of machine learning than in 2015.
- Compared to 2018, the number of vacancies for data analysis specialists increased by 1,4 times, and for machine learning - by 1,3 times.
- 38% of open vacancies are in IT companies, 29% in companies from the financial sector, and 9% in business services.
The situation is fueled by numerous online schools that train those same juniors. Basically, training takes from three to six months, during which students have time to master the main tools at a basic level: Python, SQL, data analysis, Git and Linux. The output is a classic junior: he can solve a specific problem, but he still cannot understand the problem and independently formulate the problem. However, high demand for specialists and hype around the profession often gives rise to high ambitions and salary requirements.
Unfortunately, an interview in Data Science now usually looks like this: the candidate says that he tried to use a couple of libraries, he cannot answer questions about how the algorithms work, then he asks for 200, 300, 400 thousand rubles a month in his hands .
Due to the large number of advertising slogans like βeveryone can become a data analystβ, βmaster machine learning in three months and start making a lot of moneyβ and the thirst for quick profit, a huge stream of superficial candidates poured into our field with absolutely no system training.
Victor Cantor
Chief Data Scientist at MTS
Who are employers looking for?
Any employer would like his juniors to work without constant supervision and be able to develop under the guidance of a team leader. To do this, a beginner must immediately master the necessary tools to solve current problems, and have a sufficient theoretical base to gradually offer their own solutions and approach more complex problems.
With tools for beginners on the market, everything is quite good. Short-term courses allow you to quickly master them and get to work.
According to a study by HeadHunter and Mail.ru, the most demanded skill is knowledge of Python. It is mentioned in 45% of data scientist jobs and 51% of machine learning jobs.
Employers also want data scientists to know SQL (23%), be proficient in data mining (Data Mining) (19%), mathematical statistics (11%) and be able to work with big data (10%).
Employers looking for machine learning specialists, along with knowledge of Python, expect the candidate to be proficient in C ++ (18%), SQL (15%), machine learning algorithms (13%) and Linux (11%).
But if the juniors are doing well with the tools, then their leaders face another problem. Most course graduates do not have a deep understanding of the profession, so it is difficult for a beginner to progress.
I am currently looking for machine learning specialists to join my team. At the same time, I see that often candidates have mastered individual Data Science tools, but they do not have a deep enough understanding of the theoretical foundations to create new solutions.
Emil Maharramov
Head of Computational Chemistry Services Group, Biocad
The very structure and duration of the courses does not allow you to go deep to the required level. Graduates often lack the same soft skills that are usually overlooked when reading a job posting. Well, really, who among us will say that he does not have systemic thinking or a desire to develop. However, in relation to a Data Scientist, we are talking about a deeper story. Here, in order to develop, you need a fairly strong bias in theory and science, which is possible only during long-term studies, for example, at a university.
Much depends on the person: if a student with a good base in mathematics and programming passes a three-month intensive course from strong teachers with experience of team leaders in top companies, delves into all the course materials and βabsorbs like a spongeβ, as they used to say at school, then there will be problems with such an employee later No. But 90-95% of people, in order to learn something forever, you need to learn ten times more and do it systematically for several years in a row. And this makes master's programs in data analysis a great option to get a good foundation of knowledge, with which you wonβt have to blush at the interview, and it will be much easier to do your job.
Victor Cantor
Chief Data Scientist at MTS
Where to study to find a job in Data Science
There are many good Data Science courses on the market and getting an initial education is not a problem. But it is important to understand the direction of this education. If the candidate already has a strong technical background, then intensive courses are what you need. A person will master the tools, come to the place and quickly get used to it, because he already knows how to think like a mathematician, see the problem and formulate problems. If there is no such background, then after the course there will be a good performer, but with limited opportunities for growth.
If you have a short-term goal of changing professions or finding a job in this specialty, then some systematic courses are suitable for you, which are short and quickly provide a minimum set of technical skills so that you can qualify for an entry-level position in this field.
Ivan Yamschikov
Academic Director of the Online MSc in Data Science
The problem with courses is that they give a quick, but minimal overclocking. A person literally flies into the profession and quickly reaches the ceiling. To get into the profession for a long time, you need to immediately lay a good foundation in the form of a longer-term program, for example, in a master's program.
Higher education is suitable when you understand that this area is of interest to you for the long term. You don't want to go to work as soon as possible. And you donβt want to have a career ceiling, and you also donβt want to face the problem of a lack of knowledge, skills, a lack of understanding of the general ecosystem through which innovative products are developed. This requires a higher education, which not only forms the necessary set of technical skills, but also structures your thinking in a different way and helps to form some vision of your career in the longer term.
Ivan Yamschikov
Academic Director of the Online MSc in Data Science
The absence of a career ceiling is the main advantage of the master's program. For two years, the specialist receives a powerful theoretical base. This is how the first semester of the NUST MISIS Data Science program looks like:
- Introduction to Data Science. 2 weeks.
- Fundamentals of data analysis. Data processing. 2 weeks
- Machine learning. Data preprocessing. 2 weeks
- EDA. Intelligence data analysis. 3 weeks
- Basic machine learning algorithms. P1 + P2 (6 weeks)
At the same time, you can also gain practical experience at work. Nothing prevents you from getting a junior position, once the student has mastered the necessary tools. That's just, unlike a graduate of courses, the master does not stop his education on this, but continues to delve into the profession. In the future, this allows you to develop in Data Science without restrictions.
On the website of the University of Science and Technology "MISiS" for those who want to work in Data Science. Representatives of NUST MISIS, SkillFactory, HeadHunter, Facebook, Mail.ru Group and Yandex, talk about the most important:
- How to find your place in Data Science?,
- βIs it possible to become a data scientist from scratch?β,
- βWill there be a need for data scientists in 2-5 years?β,
- βWhat tasks are data scientists working on?β,
- "How to build a career in Data Science?"
Online learning, public education diploma. accepted until 10 of August.
Source: habr.com