research

My research sits at the intersection of Generative AI, large language models, agentic systems, and trustworthy, data-centric machine learning. Below are the main themes I work on, with representative publications.

Foundation Model Virtualization & Serving

Rethinking how foundation models are managed, selected, and served at scale — including an operating-system-inspired layer that virtualizes foundation models for enterprise and scientific workloads.

• It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer — ICML 2026
• Advancing Autonomous Microscopy Agents with Domain-Guided Dynamic Retrieval in a Virtual Foundation Model OS — TPC 2025

Agentic AI Systems & Guardrails

Building reliable, policy-governed autonomous agents — with a focus on guardrails for MCP-driven infrastructure agents, reliable tool use, and automated evaluation of agent behavior.

• InfrastructureSentinel: Policy-Enforced Guardrails for Secure MCP-driven Infrastructure Agents — AAAI 2026
• Beyond Curated Benchmarking: Automated Evaluation of LLM Agents for Safe and Reliable IT Infrastructure Management — AAAI 2026 Workshop

LLM Selection, Evaluation & RAG

Data-efficient evaluation and co-optimized recommendation of LLMs for practical tasks, and retrieval-augmented generation that is both accurate and low-hallucination.

• SubLIME: Subset Selection via Rank Correlation Prediction for Data-Efficient LLM Evaluation — ACL 2025 (Best Theme Paper, Oral)
• CEBC: Conformal Evidence-Bounded Control for Low-Hallucination Vision–Language Generation — ACL 2026
• Co-optimizing Recommendation and Evaluation for LLM Selection — ICLR 2025 Workshop

Time-Series Foundation Models

Adapting multivariate time-series foundation models for forecasting and anomaly detection under limited historical data.

• Adaptive Refinement of Time Series Foundation Models via Pattern and Context-Awareness — NeurIPS 2025 Workshop

Trustworthy & Data-Centric AI

Dataset valuation and lineage, data discovery, fairness-aware data generation, and explainability for AI pipelines.

• Constructing a Metadata Knowledge Graph as an Atlas for Demystifying AI Pipeline Optimization — Frontiers in Big Data 2025
• LLM-Guided Counterfactual Data Generation for Fairer AI — WWW / ICLR 2024 Workshops

Graphs, Hypergraphs & Network Biology

My doctoral research developed algorithms for rich graph structures — hypergraph clustering, hyperedge prediction, and multilayer network centrality — with applications in molecular and network biology.

• MultiCens: Multilayer Network Centrality Measures to Uncover Molecular Mediators of Tissue–Tissue Communication — PLOS Computational Biology 2023
• Hypergraph Clustering by Iteratively Reweighted Modularity Maximization — Applied Network Science 2020
• HPRA: Hyperedge Prediction using Resource Allocation — WebSci 2020

See the publications page for the complete list, and projects for selected themes in more detail.