research

I develop autonomous AI workflows for materials discovery, design, and simulation — integrating quantum mechanics, atomistic modeling, and continuum-scale simulation with reinforcement learning, graph neural networks, and multi-agent LLM systems. My research is organized into four interconnected thrusts.

1. Agentic AI for Autonomous Scientific Simulation

The next frontier of computational materials science is not faster simulation but autonomous simulation — AI systems that can specify, execute, debug, and analyze physics-based models with minimal human intervention. My flagship project, AutoMOOSE, is a multi-agent LLM framework that automates the end-to-end workflow of phase-field simulations in MOOSE, from natural-language problem specification through HPC execution and post-processing.

Looking forward, this thrust extends to agentic DFT workflows, closed-loop materials-discovery pipelines that combine simulation with autonomous experimentation, and scientific reasoning agents that can navigate the full multiscale stack from electrons to devices.

2. Reinforcement Learning for Materials Design

Long before LLM agents, I have been treating materials discovery as a sequential decision-making problem. The CASTING framework uses continuous-action-space tree search for crystal-structure prediction and inverse design, while the BLAST toolkit applies reinforcement-learning strategies — originally inspired by AI for board games — to the construction of empirical and neural-network interatomic potentials.

Together, these methods have yielded transferable force fields for silica, silicene, arsenene, phosphorene, bismuthene, gold and nickel nanoclusters, transition metals, and high-entropy MXenes — and have been extended through multi-reward, hierarchical, and symbolic-regression variants to produce physically interpretable potentials.

3. Multiscale Modeling from Quantum to Device

I bridge length and time scales by integrating DFT, machine-learning force fields, classical molecular dynamics, and phase-field modeling within a single research program. Recent work spans:

• Neuromorphic computing materials — Mott systems and hydrogenated perovskite nickelates as reconfigurable platforms for AI hardware (Science, Science Advances, Nano Letters, ACS Nano).
• Thermal management for high-performance computing — projection-based Cahn–Hilliard–Navier–Stokes phase-field modeling of two-phase flows in microelectronics cooling systems.
• 2D and high-entropy materials — discovery and characterization of 2D MXenes, transition-metal dichalcogenides, and high-entropy alloys for tribological and electronic applications.

4. Data-Driven Materials Discovery and Infrastructure

Scalable AI for materials requires scalable data. I co-founded the Quantum Cluster Database — the largest open repository of atomically precise nanoclusters, cataloging more than 70,000 DFT-computed structures across 55 elements. This thrust also includes data-driven studies of process–structure–property relationships in additive manufacturing, machine-learned phase diagrams for accelerated phase discovery (boron, carbon, transition metals), and ML-based elastic-property prediction.

A detailed research statement is available here.