Received June 1, 2023.
Basetis is a Barcelona-based IT consulting firm comprising over 350 employees with a strong focus on social change.* For instance, we have adopted the Teal philosophy, which stands on three pillars: self-management, wholeness, and evolutionary purpose. As a result, decision-making and leadership responsibilities are distributed among Basetis personnel rather than following a rigid hierarchical structure.** Our business encompasses a variety of ICT services, including software and mobile application development, graphic design, cloud infrastructure management, as well as data analytics and artificial intelligence (AI) services.
In these latter domains, Mathematics constitutes a pivotal component in the successful execution of our projects. For this matter, Basetis employs numerous mathematicians and physicists. In fact, most of the very first members of Basetis were alumni of the FME, creating a close connection between the company and the faculty. As a result, many FME students come to Basetis every year for internships, some of them subsequently joining the company on a permanent basis, constituting a significant portion of the company.
Mathematicians at Basetis possess a deep intellectual curiosity which led the company to strategically establish its own AI Team five years ago. This decision capitalized on the intersection between Mathematics and AI, leveraging the strengths of the firm’s STEM profiles to develop cutting-edge AI solutions for our clients. Basetis already provided data analytics solutions since its foundation, so the incorporation of AI was a natural next step as its relevance increased within the IT industry.
From a mathematical perspective, AI projects usually come with intriguing challenges. There are a wide variety of clients, with different needs and data, hence our spectrum of services is rather broad. In all cases, Mathematics is the backbone of our work, as it provides the theoretical foundation for the algorithms and models that power every aspect of AI technology. These projects can be divided into three distinct categories:
Fraud detection in bank transactions
1. Advanced and Predictive Analytics
Developing accurate predictive models that can forecast future or real-time outcomes based on historical data is one of the most important challenges faced by the AI Team. Examples of such applications include automated fraud detection in bank transactions, energy consumption forecasting, and stock and demand predictions, among others. The successful execution of these projects requires a profound understanding of statistical principles and techniques. Initially, historical datasets must be subjected to rigorous statistical analysis and interpretation to identify the relevant variables and create new synthetic variables from the data. Also, a comprehensive grasp of the mathematics and complexity underlying the typical models used in machine learning is vital for effective model selection.
All models inherently involve a procedure for error minimization at their core. However, the specifics of this procedure may vary depending on the model structure, leading to different model behaviors. For instance, decision trees, gradient boosting machines, and logistic regression are examples of machine learning models with a different structure but which can solve the same task. Once the models are fitted to the data using the selected variables, it is crucial to carry out a thorough analysis of their performance. This analysis enables the identification of potential risks and biases associated with the model, enabling appropriate steps to mitigate such issues.
2. Optimization
Another key focus of the AI Team is mathematical optimization. Depending on the problem, we can differentiate between continuous optimization and discrete optimization. While continuous optimization problems can be tackled efficiently with machine learning methods, discrete optimization relies on fundamentally different techniques such as graph-based algorithms, mixed-integer programming, Monte Carlo methods, and hybrids of those.
Assigning drivers to services
For instance, the team used graph theory to solve the problem of assigning drivers to services, using one graph to parameterize optimal service concatenation through a minimum path cover problem, and another one to determine the optimal matching between drivers and routes. In such a way, the number of drivers required and the waiting times were minimized and the service efficiency was improved. On a different project, graphs were used to represent a set of academic problems that a student had to face and the dependencies between them. The optimal education itinerary for individual students could be computed using Dijkstra’s algorithm and Monte Carlo-based optimization methods
3. AI for perception tasks
In addition to these challenges, we also work on developing natural language processing (NLP) systems, computer vision (CV) algorithms, and other AI applications that require a strong mathematical foundation. Using state-of-the-art techniques, our team has deployed a wide variety of successful AI solutions.
Detecting the growth of bacterial colonies
The AI team has built a chatbot able to function as a drive-through window assistant for a fast-food restaurant chain, using NLP technology to make it capable of understanding the intention of the client’s words when listening, and guiding the conversation in order to deliver the order efficiently. On the CV front, the team has developed a system to automatically identify, classify, and quantify olives from images taken of the conveyor belt for real-time quality monitoring. Similarly, an algorithm for detecting the growth of bacterial colonies on a Petri dish was developed, being able to identify typical species and quantify their amount, which can be very helpful or early diagnosis. Moreover, our team has also worked on audio-perception solutions, for instance developing a device able to anticipate malfunctions on industrial machinery by processing sensor data and sound records of the machines.
Overall, the work of the AI team at Basetis is highly technical and requires a strong background in mathematics. By leveraging our expertise in math and AI, we are able to provide innovative solutions that help our clients to achieve their business goals.
Press headlines
- El País/Business Management:
Companies without a boss work (and very well) - Indicador d’Economia/TIC:
End hierarchies to empower talent
* Marc Castells and Víctor Roquet founded Base Technology and Information Service (Basetis) on 6 November 2009, a company in the ICT industry initially focused on providing professional services.
** Basetis is made up of a team of people who are passionate about information and communications technologies. We share an entrepreneurial spirit and base our way of doing things on trust.