RESEARCH
Machine learning that can be examined, tested, and used with confidence.
I develop machine-learning methods that combine interpretability, domain knowledge, and robust representation learning, with applications in energy systems, perception, finance, medical imaging, and human-centered AI.
Explainable & Interpretable AI
Understanding how predictive information is represented and how explanations can support human reasoning.
How can internal neural representations become useful evidence without changing the deployed model?
My work includes mechanistic analysis, concept probing, finite-difference explanation, representation visualization, and human-centered evaluation of explanations.
Physics-Guided Machine Learning
Integrating physical knowledge and system structure into data-driven models.
How can learned models respect useful domain structure while retaining predictive flexibility?
This direction focuses on model architectures that incorporate physical guidance, memory, and structured inductive biases for energy modeling and forecasting.
Time-Series & Energy Intelligence
Adaptive forecasting and robust modeling for heterogeneous real-world temporal data.
How can temporal models adapt across consumers, environments, and incomplete observations?
I study recurrent, kernel-based, physics-guided, and diffusion-based approaches for forecasting and missing-data reconstruction in energy systems.
Responsible & Applied AI
Evaluation methods that connect computational performance with human perception, understanding, and domain outcomes.
Are we measuring what people and real-world applications actually need?
This work spans perceptual image assessment, explanation modalities, internal-consistency measures, and applied AI studies across high-impact domains.
OPEN RESEARCH SOFTWARE
Research implementations that can be inspected and reused.
Linked directly to publication records and refreshed from the site’s research metadata.
RESEARCH SOFTWARE · LIVE
Latest GitHub research activity.
The four most recently updated repositories among the code linked to publications on this website.
Kolmogorov-Arnold recurrent forecasting implementation.
Code ↗GLIPSCodePerceptual quality assessment for AI-generated images.
Code ↗LinearLensCodeHuman-centered mechanistic interpretability research code.
Code ↗PgMNCodePhysics-guided memory network for building-energy modeling.
Code ↗Repository language, stars, update time, and release tags are requested from GitHub’s public API; the static cards remain usable if the API is temporarily unavailable.
PEER REVIEW SERVICE
Peer reviewer for leading IEEE, AI & energy journals.
Contributing expert peer review across industrial informatics, explainable AI, energy AI, and human–machine systems.
- IEEE Transactions on Industrial Informatics
- Energy and AI
- Energy Reports
- IEEE Transactions on Artificial Intelligence
- IEEE Transactions on Human-Machine Systems
TRENDING ON THIS SITE
Reader interest this month
See which research is attracting attention across the site.