The solution uses physics-informed neural networks (PINNs) to support diesel hydrotreating, combining experimental data and physicochemical knowledge of the process. The approach enables the development of more interpretable models capable of adapting to different operating conditions. Preliminary results indicate greater accuracy in predicting product quality variables. Thus, the technology has the potential to support the optimization and monitoring of HDT units.

To make safety incident reporting more agile and accessible in industrial environments, a digital transformation and user-centered design approach was adopted. Based on process mapping with HSE (Health, Safety, and Environment) specialists, a mobile application was prototyped for devices certified for ATEX Zone 0, facilitating reporting at the time of the occurrence and contributing to faster, data-driven risk management.

The use of Large Language Models (LLMs) and intelligent agents enhances conversational search in the oil and gas sector. The approach identifies domain-specific entities, such as wells, fields, and platforms, in structured and unstructured data. Through prompt engineering techniques, metadata enrichment, and domain rules, the system improves query accuracy and information retrieval. Consequently, search becomes more contextualized and relevant, supporting engineering activities and decision-making.

Nine Artificial Intelligence models were developed to support the optimization of diesel hydrotreating, an essential process to ensure fuel quality. Trained with experimental data from different feed profiles, the models enable predicting product properties, hydrogen consumption, and process performance across different scenarios. With this, they support the definition of operating conditions and feed selection, contributing to greater efficiency and reducing off-spec diesel production.