In the vast majority of cases, when it comes to “real” enterprise-level product or solution development, enterprise architects and global architectures and patterns immediately appear, high-level data models and concepts, attempts to cover everything and everything. A short list of languages and frameworks is formed, within which all subsequent development takes place. Everything “only in Java” or “only in C#” or ... (write in at your discretion).
Undoubtedly, this is a reflection of previous project experience, best world practices, readiness to pick up new business requests, and in general, this approach is justified. But in each particular case, such globalism at the stage of take-off of the product, at a time when much is still in a state of uncertainty, can simply bury the undertaking under itself and turn the project into another failure. Is it possible to change, simplify and improve something without losing quality?
It turns out that this is quite possible by combining classical software development with data science tools and approaches (hereinafter simply DS). How this can be achieved - we will analyze it step by step.
Not all developers can clearly imagine what data science is and what specialists do. The opinion in films and advertising that they are engaged in artificial intelligence and machine learning is far from reality. Yes, there is a lot of math in data science. A lot and quite complex. Yes, separate branches of DS used to be called more precisely - computer science and they were engaged in numerical algorithms. But this is just one of the directions.
In the corporate world, datascience specialists are more concerned with adapting ready-made approaches and algorithms for business data flows than with fundamental mathematics.
Bringing models and analytic applications to productive use is essentially very similar to software development.
