Galtea

Deploying, monitoring, and evaluating generative AI models

Institution:

Institution

Research Group:

Language Technologies

Researcher/s:

Jorge Palomar, Baybars Külebi

Galtea

Description:

Galtea introduces a platform developed by BSC-CNS for effectively deploying, monitoring, and evaluating generative AI models, providing a systematic approach to assess compliance with the latest EU regulations. The primary focus is to expedite the introduction of AI-based products to the market while maintaining adherence to the regulatory landscape.

Type of asset:

Service

Category:

Life Sciences

Problem:

In the rapidly evolving landscape of Artificial Intelligence, technological advances are outpacing regulatory frameworks, creating a critical gap in compliance. In December 2023, an interim agreement on the "Artificial Intelligence Regulation" was reached in the European Parliament, which aims to ensure that AI systems introduced in the European market respect fundamental rights and EU values. This means that companies will soon face the difficult task of ensuring that their AI models comply with these new standards, which will require extensive knowledge and compliance expertise to avoid potential legal hurdles and financial penalties.

Solution:

The technology is designed as a compliance platform with an important component of AI model visibility platform (MLOps). That is, the platform is intended not only to assist models in some of their most important stages (training, deployment, assessment, monitoring), but to ensure that each of these stages complies with the requirements of the AI Act or other laws that may come into play. This allows, on the one hand, to ensure transparency and traceability of the entire life cycle of an AI-based product and, on the other hand, to assess the safety, biases or toxicity, among others, of the final result.

Aplication areas:

Enterprises with multiple AI systems which need comprehensive compliance assessment tools that can be integrated across multiple products and departments, Startups with high-risk AI products which need accessible, efficient, and cost-effective compliance solutions that enable quick market entry without compromising on regulatory obligations or investors which need to perform transparent and reliable evaluations of AI compliance and risk to ensure investment viability and long-term returns, among others.

Novelty:

The uniqueness of our technology is that it evaluates the compliance of GenAI models, with special attention to LLMs, in a quantifiable, technical, and automatic way, and integrates elements of MLOps to facilitate its application in the life cycle of AI-based products.

Protection:

Proprietary software

Target market:

Large companies with multiple AI products/systems, startups with high-risk AI products, Investors or insurance companies analyzing product risks, LLM developers / API providers, authorized representatives, Law firms or IT/Business Consulting Firms.

Keywords:

AI compliance, AI Act, AI governance, AI risk management, MLOps, LLM observability, LLM deployment

TRL: 6

CRL: 2

BRL: 4

IPRL: 4

TmRL: 5

FRL: 2

More information

if you want to know more about this project do not hesitate to contact us

Contact us