Head of Research (AI)

Role overview

Lead research on foundation models for relational data, connecting scientific advances in graph learning with production systems for high-stakes enterprise decisions. This is a hands-on leadership role: you’ll set research direction, grow a small team, and work closely with data, product engineering, and MLOps colleagues to turn model improvements into useful capabilities.

Responsibilities

- Own the research roadmap for relational foundation models, including knowledge graphs, representation learning, and self-supervised or unsupervised methods. - Lead, mentor, and develop research scientists and engineers. - Establish a disciplined process for defining hypotheses, documenting proposals, running experiments, and making evidence-based decisions. - Design and implement graph neural networks for node-, edge-, and graph-level tasks, including multi-scale embeddings and temporal or inductive generalization. - Build reproducible training and evaluation pipelines with appropriate distributed training tools. - Maintain rigorous benchmarks and statistically sound model comparisons. - Partner with data, product engineering, and MLOps teams to move research into production.

Requirements

- Deep expertise in graph neural networks and relational learning, with practical experience on large-scale graph problems. - Experience with PyTorch and graph learning libraries such as PyTorch Geometric or DGL. - Strong software engineering skills, including testing, profiling, and building maintainable systems. - Demonstrated experience mentoring and leading technical projects or a small research/engineering team. - Professional proficiency in both Portuguese and English.

Nice to have

- Experience with distributed model training or inference. - Knowledge of self-supervised or contrastive learning, particularly for graph data. - Publications at leading AI conferences or substantial open-source contributions.

Benefits and work setup

- Full-time remote role associated with São Paulo. - Work on production research for relational prediction and decision-support systems.