Keynote Speakers

The measurement of the level of petroleum and its derivatives in tanks at terminals and refineries is used for operational control and custody transfer. The importance of this measurement lies in determining the volumes used to calculate payments for commercial transactions and the associated taxes. When automated, level measurement is performed with servo-operated gauges. In some terminals and refineries, however, level measurements are still carried out manually using tape measures.

Additionally, in many cases, petroleum and its derivatives contain a certain amount of water that must be drained before level measurement for commercial purposes. This water-level measurement is usually performed manually using a tape and water-detecting paste that changes color in the presence of water. Manual measurements have several associated uncertainties, do not provide continuous data, and expose operations to risk. Petrobras and UFES, through a cooperative agreement, are working together to develop an optical-fiber-based solution for continuous measurement of level and oil–water interface in tanks. Optical fiber sensors stand out for their numerous advantages, such as high sensitivity, miniaturization, and the ability to perform remote and distributed sensing. Their robustness allows operation in harsh environments with real-time monitoring and immunity to electromagnetic interference.These characteristics make them ideal for several fields, including biomedical, environmental, aerospace, security, defense, energy, public services, and industrial automation. Additionally, the development of intrinsically safe and chemically stable sensors enables even more sensitive and versatile projects, especially for critical applications in the oil and gas industry. Within this partnership, prototypes have been built and tested in the laboratory—first with mineral oil and fresh water, and later with crude oil, emulsions, and salt water—with good results. In the current phase of the project, installation and long-term field testing of the technology in a petroleum tank at a terminal are planned.

AI is rapidly shifting from cloud-centric to highly distributed, resource-constrained edge compute devices for embedded and IoT applications. This shift requires a fundamental rethink in AI algorithms, systems design, energy efficiency, and trust. In this keynote, I will explore emerging architectures, algorithms, and hardware–software co-design strategies that enable energy-efficient intelligence at the edge while preserving dependability and explainability. Drawing on recent advances made in these aspects using neurologically-inspired frameworks, such as Tsetlin machines, I will demonstrate how we can achieve high performance with drastically reduced computational cost and energy. I will then conclude by sharing my entrepreneurial journey of empowering industrial edge AI technology using logic based machine learning methods.