Bolsista Mestre ou Doutor – Cientista de Dados, Graph Neural Networks

This is a fully remote position, open to applicants in Brazil.

📋 Description

• Structure the relationship graph of legal entities by mapping nodes (clients) and edges (corporate links, transactions, guarantees, and suppliers).

• Develop pipelines to convert tabular data into heterogeneous graphs.

• Design and train architectures for Graph Neural Networks, including GraphSAGE, GAT, and GIN.

• Utilize PyTorch Geometric or DGL to propagate risk signals throughout the network.

• Conduct comparative experiments between GNN models and traditional models such as XGBoost and logistic regression.

• Validate the evolution of the Gini metric and ensure temporal robustness through backtesting against data leakage.

• Apply explainability techniques, such as GNNExplainer, to identify influencers within the graph.

• Ensure transparency and adherence to banking governance and compliance standards.

• Industrialize the model via FastAPI APIs and batch inference.

• Containerize the solution with Docker and Kubernetes for integration with credit decision systems.

⛳️ Requirements

• Completed master's or doctoral degree.

• Background in Physics, Mathematics, Computer Science, Statistics, Data Science, or related Engineering fields.

• Experience in machine learning and deep learning research.

• Proficiency in graphs and deep learning applications involving graphs, particularly GNNs.

• Knowledge of the Python programming language.

• Publications in deep learning, graphs, CNNs, and GNNs are preferred.

• Research experience involving large data volumes is desirable.

• Research experience with multidimensional edge graphs is desirable.

🏝️ Benefits

• Stipend of R$ 7,000.00 for completed master's degree or R$ 9,000.00 for completed doctoral degree.

• Remote working format.

• Availability of 40 hours per week.

• Duration of 12 months.