PRAVIA (ComPetividad industRial basada en la gestión del conocimiento de la EmpresA Vía Inteligencia Artificial)

PRAVIA

industrial competitiveness via artificial intelligence

The PRAVIA project develops an advanced conversational system based on Generative Artificial Intelligence and Natural Language Processing (NLP). Its objective is to transform the way employees access knowledge and corporate information, which is usually scattered across different systems (ERP, CRM) and formats. With PRAVIA, users will be able to make queries in natural language to access critical data in an agile and precise way, optimising decision-making, improving competitiveness and facilitating knowledge transfer among employees.

Challenges

Integrate and standardise large volumes of dispersed and unstructured data (IT systems and office software), creating a robust architecture that serves as a centralised base of business knowledge.

Implement and fine-tune (Fine-Tuning) Advanced Language Models (LLMs) such as GPT-4 or LLaMA, adapting them to the vocabulary and specific processes of each industrial environment using Retrieval-Augmented Generation (RAG) techniques.

Automate workflows and apply document understanding techniques for critical uses such as: cost control, shift planning, and retention of lessons learned in the face of generational turnover.

Solution

Development of an intelligent platform that unifies knowledge management through generative AI. It combines process mining, extraction of structured data from heterogeneous documentation (Document Understanding) and real-time inference capabilities.

The tool acts as an interactive conversational repository to optimise operational efficiency, being trained and validated in real use cases from three industrial giants, ensuring its robustness to be scaled.