Kévin Osanlou

AI Researcher Google Scholar →

Machine Learning, Large Language Models (LLMs), Planning, Scheduling & Combinatorial Search

Since 2017 I have combined machine learning with planning, scheduling and combinatorial search, putting graph neural networks inside exact solvers to make them substantially faster without giving up optimality. The work is published at AAAI and IROS, and was built for real autonomous-systems problems at Safran and NASA Ames.

As AI Research Scientist & Consultant at Talan, I now work increasingly with large language models: designing and deploying retrieval, fine-tuning and multi-agent systems in production for national infrastructure operators.

PhD in graph machine learning and automated planning, Paris-Dauphine / PSL. AAAI program committee reviewer since 2022.

Selected work All publications →

Solving Disjunctive Temporal Networks with Uncertainty under Restricted Time-Based Controllability Using Tree Search and Graph Neural Networks

Osanlou, K., Frank, J., Bursuc, A., Cazenave, T., Jacopin, E., Guettier, C., & Benton, J.

AAAI 2022

Main track

AAAI arXiv

Optimal Solving of Constrained Path-Planning Problems with Graph Convolutional Networks and Optimized Tree Search

Osanlou, K., Bursuc, A., Guettier, C., Cazenave, T., & Jacopin, E.

IROS 2019

Main track

IEEE arXiv

SWOT-based Simulation of River Discharge with Temporal Graph Neural Networks

Osanlou, K., Getirana, A., Holmes, T., & Cazenave, T.

NeurIPS 2024

D3S3 workshop

OpenReview