Published in Applied Energy, 2026
A robust planning framework that co-optimizes hydrogen-based microgrid capacity and the daily service window for remote high-power operations.
Published online in Journal of Optical Communications and Networking, 2026
A leaf-centric logical topology design that avoids routing polarization in OCS-based GPU clusters.
Published in Energy Reports, 2023
A data generation method for power systems that accounts for geographical correlations.
Accepted by NSDI '27, 2026
A learning-based approach that advances quality, scalability, and flexibility in collective communication.
Published in NSDI '26, 2026
This paper proposes a fast solver-free algorithm to address TE problems in large-scale DCNs.
Published in NSDI '26, 2026
Geminet learns a topology-agnostic update operator enabling lightweight traffic engineering under changing network topologies.
Accepted by INFOCOM '26, 2025
This paper presents ATRO, a fast algorithm designed for topology engineering in reconfigurable DCNs.
Available on arXiv, 2026
NeuroRisk is a physics-informed deep unrolled optimizer for risk-aware traffic engineering under correlated WAN failures.