Research Scientist, Relational Foundation Models
Core
Researching and applying relational foundation models for enterprise decision-making in Brazil, focusing on graph-native models for structured prediction problems like credit and fraud.
Role type
Senior applied research scientist (relational foundation models)
Builds
Graph-native foundation models for credit, fraud, growth, and monitoring decisions
Domain
Financial services / Graph Machine Learning
Deliverable
production ML models
Required skills
Graph Machine Learning, GNNs, graph transformers, representation learning, distributed training, experimental design, large-scale data handling
Preferred skills
Heterogeneous/temporal graph libraries (PyG, DGL), Ray for distributed training, Rust/C++/CUDA/Triton, JAX, production model deployment
Technologies
PyTorch, Ray, PyG, DGL, Rust, C++, CUDA, Triton, JAX
Responsibilities
Evolve model thesis and architecture for relational foundation models, design training objectives (contrastive, generative, supervised), implement transfer learning and fine-tuning strategies, conduct rigorous evaluation with temporal validation and leakage checks, build large-scale training infrastructure, optimize performance-sensitive ML systems
Seniority
Senior, hands-on IC
