Bounded Multiscale Forecasting with Hierarchical and Sequential Mesh Refinement of Graph Neural Networks
karthikLab
| Adjunct Professor, KAIST |
AI for complex systems: learning across scales, connecting scientific knowledge, and acting under uncertainty.
Research themes: AI for complex systems · Scientific ML · Earth system science · Energy markets · Reinforcement learning
All publications · Climate · Energy · Earth
Graph learning across scales
Architecture diagrams from KDD, IJCAI and WSDM. Select a figure to open it at full resolution.
Addressing Downward Memory Loss in Hierarchical GNN Forecasters Through Memory-Buffered Decoding
HoGA: Higher-Order Graph Attention via Diversity-Aware k-Hop Sampling
Learning and control
Position: Train Robustly and Certify at Runtime Under Partial Observability
Model-based reinforcement learning for ultrasound-driven autonomous microrobots
AI-driven autonomous microrobots for targeted medicine
Earlier preprint
Reducing Smoothness with Expressive Memory Enhanced Hierarchical Graph Neural Networks. Thomas Bailie, Yun Sing Koh, S. Karthik Mukkavilli and Varvara Vetrova (2025).
Graduates, PhDs, Postdocs
Please apply directly to KAIST, then indicate interest in AI co-supervision with me alongside a KAIST professor focusing on a domain.