Along with the product, RnD prepares technical documentation:

[rifat28dddd]

Data analysis. First of all, the RnD specialist will collect and analyze all the initial data: business requirements, available datasets, and feedback from clinical specialists. This will help to create a complete picture and detect possible problems.
Gathering additional information. RnD will then study the scientific literature on developments in AI in medicine and NLP technologies. The specialist does not have to know everything, but must be able to quickly find the necessary information.
Proposing and testing hypotheses. At this stage, RnD formulates how a specific problem can be solved and organizes the preparation of a dataset for testing hypotheses.
In some teams, the R&D specialist creates the MVP (minimum viable product) on their own . This allows them to test the hypothesis without involving the development team and to abandon an unsuccessful idea in time.

Transferring solutions to the product team. If the mexico telegram data engineer can prove on test data that his hypothesis worked, and this is confirmed by clinical specialists, then the project can be transferred to the product team. There, developers will work on the code, and designers will create a beautiful and user-friendly interface for the service.
When preparing hypotheses, RnD interacts with the business department and relevant specialists. It is not necessary to be both a doctor and an NLP engineer to do RnD in medicine, but it is important to receive feedback from doctors and ask them questions. RnD transfers ready-made solutions, for example, to the backend team, for implementation in the product.


Vladimir Vorobyov,
Sber analyst

tech debt - a list of problems in the software code that need to be fixed, otherwise they will require additional costs and resources in the future;
current and target architecture - how the project is implemented now and how it should look in the future;
ADR (Architecture Decision Records) is a document that describes how and why certain architectural decisions were made during the development of a project. This will allow developers to avoid repeating the mistakes that the RnD team has already gone through;