Research
Computational Discovery Across Quantum and Photonic Systems
Our work focuses on discovering compact, efficient, and interpretable representations of complex scientific systems.
AI-Assisted Quantum Algorithm Discovery
A central direction of my current research asks whether artificial intelligence can help reveal the structures and principles that underlie efficient quantum algorithms. Rather than treating circuit design only as a manual process, this work investigates computational frameworks that can search, analyze, reduce, and interpret candidate quantum circuits.
Topics include:
- AI-assisted quantum algorithm discovery
- Automated quantum circuit structure search
- Compact quantum representations
- Hybrid quantum-classical optimization
- Quantum circuit reduction and resource analysis
- Identification of transferable quantum design principles
Quantum Simulation
I investigate quantum algorithms for physical and chemical systems with particular emphasis on reproducibility, resource efficiency, and physically meaningful model reduction.
Current directions include:
- Quantum computational chemistry
- Variational quantum eigensolvers
- Atomic and molecular Hamiltonians
- Photonic and quantum-material Hamiltonians
- Reduced active-space representations
- Quantum resource estimation
- Robustness across molecular geometry and model choices
A key objective is to determine how much circuit and parameter complexity is necessary to reproduce the behavior of a target physical system.
Artificial Intelligence for Nanophotonics
Nanophotonic design problems often require repeated electromagnetic simulations over high-dimensional parameter spaces. My research uses machine learning to accelerate forward modeling and inverse design while preserving physically meaningful behavior.
Research topics include:
- Multilayer optical structures
- Metasurfaces
- Optical gratings
- Photonic and optical sensors
- Spectral-response prediction
- Surrogate modeling
- High-dimensional inverse design
- Physics-guided learning
Electromagnetic simulations and analytical models are used to generate datasets that support systematic machine-learning studies.
Explainable and Efficient Scientific Machine Learning
High predictive accuracy is not enough when a scientific model is used to reason about a physical system. I am interested in identifying what information neural networks learn, which variables matter most, and how model complexity relates to the intrinsic complexity of the underlying dataset.
Methods include:
- Explainable AI
- SHAP-based feature importance
- Neural-network pruning
- Model compression
- Dimensionality reduction
- Autoencoder latent-space analysis
- Intrinsic dimensionality estimation
- Bias-variance and capacity analysis
Quantum Cybersecurity
I am also developing research at the intersection of quantum information and cybersecurity, especially for cyber-physical systems in which unpredictability, adaptation, and adversarial learning are central concerns.
Topics include:
- Quantum-enhanced cybersecurity
- Quantum-generated entropy
- Cyber deception
- Quantum honeypots
- Moving-target defense
- Learning adversaries
- Hybrid quantum-classical cyber systems