Quantum computing and machine learning

The group investigates the application of quantum algorithms to well data analysis and segmentation problems. The approach includes the computational representation of the data and the simulation of quantum operations on classical computers, making it possible to explore new formulations and compare them with conventional machine learning methods.

The work includes feasibility studies, prototype implementation and evaluation of computational cost and scalability. The connection with problems already studied by the group provides a basis for examining the quality of the representations and the relevance of the methods for oil and gas applications.

This front is at an exploratory stage, with proofs of concept in classical simulation. The investigation considers both the methodological potential and the execution and training limitations, without assuming demonstrated quantum advantage or a validated industrial solution.

Data engineering and model generalization

Preparing scientific data is a central part of the group's expertise: pre-processing, normalization, building training sets and defining validation strategies are part of model development. The goal is to preserve the information relevant to the physical problem and to evaluate how methods behave under conditions different from those used in training.

In well data, generalization across wells is an important methodological question. The group evaluates architectures and test procedures to understand the influence of differences between datasets, identify limitations and guide the application of models to new interpretation contexts.

Simulation and high-performance computing

The integration of physical modeling, numerical simulation and machine learning supports synthetic data generation, network training and hypothesis comparison. This capacity makes it possible to investigate relationships between signals, images and physical properties, and to build controlled scenarios for evaluating methods.

High-performance computing, including GPU resources, supports intensive computational experiments and the development of approaches at different scales. Execution strategies take into account training cost, data volume and the requirements of each scientific problem.

Scientific software and method integration

The group develops scientific code and organizes routines for data preparation, training, inference and evaluation. Implementing the methods connects the study of algorithms to building tools that can be examined, improved and integrated into research workflows.

The distinction between development and execution environments, code organization and the comparison of computational alternatives support research continuity and knowledge transfer. These tools mature alongside the scientific evaluation and the application conditions of each method.