Mercury, BBVA’s Python library is available to all machine learning and data science developers

Developing products based on artificial intelligence (AI) and machine learning (ML) involves a significant investment of time and money, from the design stage to the implementation process of data algorithms. That’s why the team of over 900 data scientists at BBVA I decided to create a library open source baptized in the name of Mercury.

thanks for the BBVA Mercury Library It is possible to streamline processes, avoid duplication, promote globalization and reuse of analytical products. All these codes are shared on the co-op platform github So that external programmers can Driving innovation in financial technology In corporate and investment banking.

a team BBVA Advanced Data Analytics It works jointly in its offices in Spain, Mexico and Latin America to create a common code that makes it possible to eliminate the complexity of algorithms, ensuring compliance with current regulations and design recommendation systems. In addition, it promotes new processes such as Data labeling process In which the bank adds new categories of expenses in its application.

maturana workwho is responsible for the specialization of large language models and mercurial models at BBVA, stresses that this kind of algorithm is not very frequent in the “open source” field, which is Advances in financial technology innovation. However, the key to Mercury’s success lies in contributions that other developers are implementing, allowing their algorithms to continue gaining traction. BBVA uses other “open source” libraries to solve cases, such as sparkAnd tensioner flow also scikit-learn.

Mercury core

Mercury was born in 2019 BBVA AI Factory. It is organized in Multiple small warehousesin a way that follows a modular design in which each one is independent and has its own More than 300,000 lines of code. All the algorithms you host must meet a series of requirements, especially since they can be used jointly by different teams of developers. Likewise, the code must be of high quality, tested, and highly functional.

Another benefit that Mercury offers data scientists in different financial entities is Force measurement. With this powerful tool, analysts will be able to calculate the power of ML models and reduce the risk of issuing erroneous data, providing greater quality in highly variable environments.

Mercury’s initial release had limited features, but it still managed to gain weight in the AI ​​factory as data scientists shepherded the integration process. Component reuse within your projects. In fact, the BBVA team at Mercury used all of the components developed with X programan effective data experimentation device.

Mercury structure

Mercury’s library is modular, allowing users to install only the parts they need. the Small, self-contained packages which are divided into:

• Datagram: Utility package that automatically intervenes and calculates various stats. Checks if different datasets match the same schema for drift calculation.
• Interpretability of Mercury: It presents methods and techniques for interpreting ML models, both locally and globally, providing a better understanding of how the ML model works.
• Mercury control: It is a package for monitoring models and their performance in production. In this way, changes in the data distribution can be detected and the accuracy in time of overlapping can be estimated.
• Mercury rollers: Analyze sequences of events extracted from transactional data, whether generated manually or automatically.
• Mercury resistance: It is defined as a lightweight framework capable of running robust tests on ML models and datasets, ensuring that flows are robust against certain conditions (label leaks or input datagram issues).
• Mercury Cetri: It is defined as a C++ library for creating, updating, and querying objects. Sitrithat is, efficient queries for subsets of data.

Mercury’s user community continues to grow and so does Used in more than a third of BBVA’s advanced data analysis. During 2023, these numbers are expected to double, which indicates the promising future of this efficient, open and collaborative platform for financial technology environments.