Unveiling the AIaaS paradigm

raquel lazcano

Raquel Lazcano López

R&D Team Leader

Mission

Our mission is to contribute to the digital transformation of Eviden BDS customers through the development of innovative services and applications based on value-driven trustworthy Artificial Intelligence.

Image
Silhoutte of virtual human

Our research roadmap is guided by a human-centric, ethical and trustworthy approach that is fully aligned with fundamental European values. We aim also to facilitate access to specialized hardware and software resources and their management, creating and adapting new tools and platforms able to use modern paradigms leveraging the computing continuum to produce more efficient and greener applications. These new tools and platforms allow us to unveil the Artificial Intelligence as a Service (AIaaS) and Machine Learning Operations (MLOps) paradigms, which abstract the complexity of distributed and heterogeneous infrastructures.

However, Machine Learning and analytics require data, and nowadays it is not possible anymore to rely just on the information gathered from a single client, and clients are usually unwilling to share their own data with others, since now, more than ever, data is becoming the new currency. Hence, there is a real need to create not only trustworthy Artificial Intelligence, but also to grant the possibility of doing it in a federated manner, so data privacy is always ensured.

Assets and products

The team is working on the following key assets and products:

  • Fast Machine Learning Engine (FastML): A Machine Learning toolbox for job orchestration on High Performance Computing (HPC) nodes, hiding the complexity of jobs management from the user perspective.
  • A set of microservices for optimizing HPC nodes resources, leveraging Artificial Intelligence to improve energy efficiency and decarbonization processes.
  • A Federated Learning (FL) framework to guarantee data privacy preservation while leveraging the collective intelligence of multiple clients.
  • End-to-end data management, data analytics and AIaaS platforms that cover the complete life cycle of Machine Learning models (MLOps) to guarantee trustworthiness, reproducibility, robustness and energy efficiency.