The scope of Data Science is very dynamic. The digital industry remains one of the few stable areas that continues to grow and strengthen its position in the market. That is why Big Data specialists or data analysts (data scientists) are especially valued today. Not surprisingly, there is a shortage of qualified specialists in the market. How to solve the staffing issue? And where to look for a person who can lead your project to success? This is what we will talk about today. Briefly about the essence of concepts.
The term has a broad interpretation, and it can be called "basis". In other words, this is a field of knowledge that serves as the “basis” and covered many related sciences: statistics, mathematics, computer science, machine learning, etc. There are also more “business” interpretations, for example: Data Science is an art or a way of transforming data into concrete solutions.
It can be a combination of such disciplines to perform actions for: machine learning, artificial intelligence, neural networks, predictive analytics, advanced analytics, data mining, big data.
A data analyst is a specialist who analyzes large amounts of data using specially designed algorithms in order to identify patterns and relationships invisible to the naked eye. If he manages to create the right formula, research can lead not only to the task for which he was hired, but also to scientific discoveries in fields ranging from sociology to mathematics. A specialist in this field, like herself, does not have a clear definition and terms of reference.
When a person calls himself a "data-scientist", it can be difficult to understand what part of this wide range of tasks he is engaged in. To find out exactly what he specializes in, it will turn out only through personal communication.
In a narrow environment, they joke: a data scientist is a person who knows statistics better than a programmer, and programs better than statistics. As you know, in every joke there is some truth, and, in fact, this definition is quite close to the truth.
All this is good, but how can a manager find “that very” data scientist? The first option is a classic one: post a vacancy on job search sites or view the CVs of applicants on your own (the downside is that the CVs of analysts are mixed there with the CVs of engineers, statisticians, programmers, etc.). The second is more modern: try to find a specialist in social networks, for example, in one of the profile groups on Facebook.
By the way, recently the administrators of one of these communities collected statistics about their users and identified four types of data scientists:
Researchers: people who specialize in statistics, have little knowledge of programming and are poorly oriented in business issues.
Developers: people who are good at programming but don't understand statistics and mostly deal with the technical aspects of the analytics process.
Creatives: people who can do everything are like free-flying birds;
Entrepreneurs: professionals who, along with technical skills, are well versed in business processes.
The problem, as in many other areas, is that there are few highly qualified specialists. As a feature, one can single out the fact that it is very easy to interview them - we will talk about portfolios and completed projects. At the same time, it is difficult for both sides to “hang noodles on their ears”.
On the contrary, there are a lot of junior specialists and those who want to become them, including people who have taken a couple of courses and demand a salary almost like in Silicon Valley. And it is just the most difficult to hire just such employees. Indeed, in the future, we want them to grow, become more independent and better versed in the field.
You can hire 10 people with red diplomas from Phystech, but it is impossible to guess which of them will be gifted enough to “drag” a project on their own. These are complex psychological nuances that have to be identified in the process of communicating with a person.
There are also quite a lot of middle-class specialists, they can be found for a reasonable number of meetings with candidates, weeding out beginners who overestimate themselves. But high-class experts cannot be found with a simple search.
If you post such a vacancy on any recruiting site, many inadequate personalities will respond, none of which you will want to invite to your office for a conversation, ”Natekin believes. You have to look for the right specialists either with the help of professional headhunters, or through personal connections and communities. The rest can be found in open sources and even through academic connections and events. However, this applies to many categories of professionals.
The most effective way to find top-level professionals is to ask their peers for recommendations. They especially like it when they can directly communicate with people from the team of a potential employer. This helps to get to know each other in advance and evaluate your prospects. Thus, personal connections are the fastest and most effective way to hire the right people. The most important thing in interviewing those who call themselves data scientists is to talk as much as possible about their practical experience and completed projects. Be interested in the reasons for choosing certain approaches and alternative solutions to the problem.
There are a large number of "tricky" interview questions, the correct answers to which can be simply learned. The theory of interest to the company can also be memorized in the process of walking through companies. But it is extremely difficult to fabricate a portfolio, if there are projects, they are visible. Moreover, as your luggage of the work done, you can show not only commercial projects that you did for the company. Works from hackathons and competitions, and even home projects, are quite suitable. All this is a clear sign of having a certain experience behind them.
Alexey Natekin, founder of Data Mining Labs: If you want candidates don't in interviews: don't lie in job openings. For example, data cleaning is a routine but necessary procedure for any project. This should be honestly said in the vacancy, not limited to attractive creative duties.
When choosing tools, you need to focus on open platforms. They are extremely popular among developers, and you should not trust the commissioned research by Gartner, which names well-known commercial products among the leaders. The choice of specialists is not accidental: development on free products is convenient and no less qualitative than on proprietary solutions. And in no case should you explain to a specialist how he should work, especially if he himself is far from this area.
McKinsey analysts back in 2012 predicted a huge shortage of data scientists, which in the US alone by 2018 should have been between 140,000 and 190,000 people. This prediction was often cited, but no one paid attention to the next paragraph of the same report, which says that there will be a lack of 1.5 million managers who can ask analysts the correct questions.
Interesting results were shown by a survey of employers about specialists in the processing of big data (Big Data). So, the main demand for Big Data analysts is formed by:
¤ IT players;
¤ Telecom-companies;
¤ Banks;
¤ Large retail chains;
The highest dynamics in the use of big data is shown by:
¤ Banking sector;
¤ Public administrations;
¤ Agriculture;
However, according to the survey, data scientists only work in 6% of companies. This can be explained by the lack of a common terminology for designating a big data specialist in the Russian data market. Jobs that imply an identical specialization may have a variety of titles. Among the most popular:
¤ big data analyst;
¤ mathematician/mathematician-programmer;
¤ system analysis manager;
¤ big data architect;
¤ business analyst;
¤ BI analyst;
¤ information analyst;
¤ data mining specialist;
¤ machine learning engineer;
Despite the large number of names, Big Data specialists are usually classified into 2 large groups: Analysts (data analysis, pattern detection and model building). Engineers (storage transformation and data access).
The best store of knowledge and skills for working in the field of big data can be obtained at higher educational institutions in the following areas:
"Applied math";
"Informatics";
"Math statistics";
At the same time, it is worth paying special attention to the methods of mathematical statistics, algorithms for data analysis and mathematical modeling, modern technologies for processing big data, the basics of working with relational databases and the SQL language. Knowledge of English at the level of reading technical documentation will significantly increase not only your competitiveness, but also your potential salary.
Depending on the area in which the data scientist works, the list of his duties may include:
¤ organization of the data collection process for the purpose of subsequent operational processing;
¤ analysis and forecasting of consumer behavior;
¤ customer base segmentation (clustering, classification, modeling, forecasting);
¤ personalization of product offers;
¤ analysis of the effectiveness of internal processes and operations and their optimization;
¤ risk analysis;
¤ detection of fraud based on the study of suspicious transactions;
¤ integration of data from different sources (multi-channel sales, marketing, Internet);
¤ formation of periodic reports for evaluation of results, visualization and presentation of data.
The very name of the data scientist profession suggests that a data specialist must, first of all, be a scientist, a person with an analytical mindset and the ability to draw reasonable conclusions from incredible amounts of information in the shortest possible lines. Scrupulous, attentive, precise - most often he is both a programmer and a mathematician. In addition, such a specialist must be business-oriented, because the entire process of his work is ultimately aimed at solving a business problem. The advantage will be the ability to clearly formulate your thoughts, which will help to present the final product of your work to the employer and raise its price.
On the agenda is a list of questions that need to be asked in order to recognize the candidate as a potentially successful and useful data scientist. Last month we talked about how to prepare for an interview with a data scientist, and today we will present a view from the “opposite side of the negotiating table”. So, what questions, other than the obvious “What do you know about our company?” and “What brings you to us?” should I ask? Your attention - TOP-10 questions for an interview with a data analyst.
1) A variation on one of the most typical questions in any job interview - with a practical twist. It allows, among other things, to understand the candidate's interest not just in a "prestigious position in a prestigious company", but specifically in this vacancy.
2) It should be noted that this includes both real and educational projects. You, as an employer, must take into account that the direction is relatively new, so it would be short-sighted to dismiss the "green" graduates/postgraduates/candidates. It is young people with brains that are the most profitable investment today.
3) This question, as practice shows, causes some difficulties for the candidate, but it is he who shows the level of "constructiveness" of the applicant's self-criticism, his ability to admit and analyze his mistakes.
4) The “trick question”, however, has a clear goal (albeit a double one): to determine the degree of “love” of a candidate in Data Science (however, yesterday’s student may be confused - speaking of a “trick” - then you can simply ask him to describe any algorithm, for example, machine learning) and his ability to clearly express his thoughts in front of an audience, albeit from a couple of people.
5) Did Candidate lost the thought? This, of course, still does not say anything; give him a couple of hints and see how he reacts to this "teamwork simulation". Here it is important to understand whether a person is able to adequately think and accept help/opinion from outside. By the way, many experts recommend specifically giving the task of "increased complexity", which, of course, is accompanied by hints a little later.
6) If in the previous question we check the practical skills and sociability of the candidate, then here we evaluate him in not so much theoretical training, but rather how his brain “works”. Yesterday's student is almost 100% likely to become nervous, remembering all the horrors of the session; to be honest, even an experienced “interview walker” can get confused. It is important here to make it clear that you are, as they say, "also a person" and try to create a more relaxed atmosphere so that a person can open up. It is not forbidden (rather welcome) to joke - within reason, of course.
7) Allowing the candidate to imagine for a while that he is already working for you, you create psychologically more comfortable conditions for him here and now, at the interview. However, the essence of the question is not so much to let the applicant “dream”, but to look at his expectations and, if he has any questions (which, by the way, is always good), answer them.
8) And again the question, the answer to which will show how great the experience of the candidate is not in words, but in deeds. Plus, you once again give him the opportunity to speak out - and enter into a dialogue with you.
9) An important question to ask any applicant in the field of modern technology. Remember that practical knowledge in this area becomes obsolete every two years, so the ability and desire to "keep abreast" - as a result of the motto "study, study and study again" - is an important quality for a data scientist in particular.
10) Do not be surprised if they answer you: “The one you just asked!” In addition to jokes, do not forget to “dilute” the interview with questions that are not directly related to work. Leave as many opportunities, as possible for the candidate to show their best side. This will help them make the right choice without missing out on something (or someone) important.
Categories: Data-science, Data-scientist Tags: #questions_interview, #datascience_interview