XMANAI partner spotlight - Tyris AI
Name: Dr. David Monzo
Job Title: Director of AI
Organization: Tyris AI:
Bio: Dr. David Monzo is the technical director and main investigator at Tyris AI. He obtained his PhD in Telecommunications, within the topic of Pattern Recognition and Computer Vision from Universitat Politècnica de València (UPV), Spain. During the last decade he has been working as a technological counselor and has founded several technological companies, mainly focused on Artificial Intelligence applied to optimize industrial processes.
Name: Dr. Gonzalo Safont
Job title: Data Scientist
Organisation: Tyris AI
Bio: Dr. Gonzalo Safont is a researcher and data scientist at Tyris AI. He obtained his PhD in Telecommunications Engineering from Universitat Politècnica de València (UPV), Spain, where he worked as a researcher for more than 10 years. His main research interests are non-destructive technologies, biomedical systems, and big data processing for financial applications.
Gonzalo joined Tyris.AI in 2020. Since then, he has worked on a wide variety of machine learning projects, such as ML-based automatic scheduling, predictive maintenance, and data mining.
Name: Sergi Pérez
Job title: Machine Learning Engineer
Organisation: Tyris AI
Bio: Sergi Pérez obtained the Telecommunications degree and the Master’s degree in Telecommunications Engineering from Universitat Politècnica de València (UPV), Spain, in 2016 and 2018 respectively. Since then, he has worked as a Machine Learning Engineer in different companies, joining Tyris AI in 2021, where he has worked on different types of projects related to predictive maintenance or quality control, among others. He is also pursuing a PhD focused on Explainable Artificial Intelligence (XAI) jointly between Tyris AI and Universitat de València. His research interests include XAI, anomaly detection and machine listening.
Q: What is your organisation’s role in XMANAI?
A: Tyris AI is working on different work packages within the XMANAI project, putting more effort in those tasks where our technical expertise can be useful. Among all these work packages, Tyris AI is leading the work package called “Novel Artificial Intelligence Algorithms for Industrial Data Insights Generation”.
It is in charge of providing different Artificial Intelligence algorithms to cover the needs of the manufacturing field. Following the XMANAI principles, the developed models have to provide not only accurate predictions but also a certain level of explainability. To achieve this, this set of tasks will be led by data scientists, who will investigate from different types of explainable AI (XAI) algorithms, such as graph and hybrid machine learning models, to how these could be evaluated by the demonstrators’ business experts.
Q: How does XMANAI relate with your or your team’s background and interests?
A: Our expertise in artificial intelligence and predictive analytics for industry fits perfectly within the scope of XMANAI. From an industrial point of view, Tyris AI has developed different solutions for different sectors, such as automotive, construction or metallurgy, among others. From a technical point of view, these works fall into different fields, such as predictive maintenance, quality control or supply chain optimisation.
Both the industries and the problems we have addressed suit the different use cases of XMANAI. Through a team of different engineers specialised in artificial intelligence, Tyris AI will help in the development of the different modules related to data engineering and machine learning.
Q: What is the novelty of XMANAI and the main benefits envisaged for your organisation?
A: XMANAI aims to provide an additional layer to the predictions made by black-box models. This layer will add a novel component to the traditional artificial intelligence solutions offered today through explainability so that business experts in the industry can interpret and trust the predictions made by the platform and use them to make better decisions.
The development of different algorithms that integrate explanations into the decisions they make will greatly improve our solutions and it is expected that this project will be a starting point for new projects based on the principles of explainable AI.
Q: Which target groups can benefit from XMANAI?
A: Different groups can benefit from this project. From the point of view of experts in the manufacturing industry, having a comprehensive platform that they can use to help them in their decision-making process and even understand why the platform gives a certain result is a very high added value compared to other conventional platforms.
From a data scientist’s point of view, being able to leverage a platform with a set of different tested explainable algorithms to solve different manufacturing problems is an advantage. Moreover, being able to reuse them to adapt them to new situations or to train a new one from a set of baseline algorithms and not develop it from scratch is another great advantage.
Q: As a technical partner, how do you envision the XMANAI progress beyond the state of the art?
A: In recent years, predictive systems based on Artificial Intelligence algorithms have been successfully applied to a large variety of industrial scenarios to optimize the overall production. These scenarios include intelligent production planning, predictive quality and maintenance, or product demand forecasting, among others. Despite the advantages given by such kind of solutions, still a wide range of companies refuse to apply them, mainly because of two causes: 1) the lack of understanding on how AI systems operate creates untrustful behaviors and 2) the lack of AI-trained personnel within the factories which hinders, or even makes impossible, the inclusion and adoption of AI solutions to every specific use case scenario.
The XMANAI project was born with these barriers in mind, as there is no current solution that addresses them in a satisfactory way, being powerful enough to compete with the most accurate data-based predictive solutions while enabling a flexible paradigm that can be extended to almost any industrial optimization process.
Therefore, XMANAI tackles directly the before mentioned barriers with a novel approach, providing: 1) a smart platform designed to be used by both AI-trained and AI-agnostic users, depending on the needs of each of these groups; 2) a catalogue of AI models within the platform (both baseline and pre-trained) to fit as a solution to specific current industrial problems; 3) a layer of understandability on top of the predictive models, providing rational on how the AI systems work and thus helping to avoid unwanted biases and generating confidence on the end users; 4) a set of ready-to-go interfaces to handle both AI predictions and their correspondent explanations and 5) an open framework to adapt the models to cover as many industrial scenarios as possible.