We talked honestly about data science. Today, when data science, machine learning and artificial intelligence are all around, there seem to be very different opinions about what data science is. Engineers see this as one thing, the Ukrainian academic community that has recently joined the race, practicing data scientists as third.
Let’s say there is a perception that data science is the same as statistics, which statistics departments have been doing for decades.
- There is an opinion that data science is about the big data platform and tools.
- There is – that this is the same as machine learning research, and its main task is to come up with a new one or an algorithm to solve ML in the most elegant way.
- There is – that data science is like a competition on Kaggle, and the main task is to squeeze as much accuracy out of the data as possible.
Data science is generally about purely research activities.
Each of these definitions generates a corresponding data scientific image. What should he know and be able to:
- Statistics;
- machine learning and deep learning;
- big data;
- coding;
- energy;
- project management;
- DevOps;
- communication with the customer.
From time to time, at conferences and meetups, arguments flare up about who is more right.
In this article I will try to talk about data science from the point of view – the science of data that we have been doing and continue to do with outsourcing companies, and whose goal is ultimately to bring this company profit.
Our customers. We experiment with models, services, approaches and see what we get. If we manage to find projects, if the customer is happy with the result, this is the right data science. If not – probably not very correct.
Data Science is now being added as an Engineer.
However, more often than not, such a model does not withstand strength tests. If a business analyst is not qualified in data science, the DS model may look like black magic to him, and the problem, the solution to which is simple and obvious, is exactly the same as the problem, the solution to which does not exist today. This leads to problems in communication with the team and the formation of incorrect expectations from the customer.
Further: if a data scientist develops only an algorithm, and the product itself is written by an engineer, the complexities of communication and operations. In all cases, without exception from my personal practice and the practice of ELEKS, if the model did have to be rewritten by an engineer (say, to work on a mobile platform), it ended up with problems and problems, who is to blame for the fact that the final system does not work as accurately as the prototype data science, not so fast, not for all cases, and so on.
Are outstaffing and outsourcing the same thing?
No! Many people mistakenly believe that outsourcing and outstaffing are interchangeable terms. This is most likely a common misconception due to the fact that both methods involve the process of assigning a client specific work to freelancers. Moreover, some people consider outstaffing to be one of the subtypes of the outsource data science services. To avoid ambiguity, it is necessary to consider each management method separately and in comparison with one another. And we’ll start with the definitions of each model.