I develop rigorous methods for mechanistic interpretability, human-centered explainable AI, reliable temporal learning, and evaluation of generative systems.
A secure daily workflow refreshes citation counts, h-index, i10-index, article-level citations, and the citation history from my public Google Scholar profile. The last verified snapshot remains visible if a refresh is temporarily unavailable.
Machine Learning with Applications · 2026
Studies explanation modalities through the outcomes that matter to people: trust calibration, comprehension, and decisions.
View research record
My research moves from the internal mechanics of neural models to the human consequences of their explanations and predictions.
Tracing concepts, representations, and predictive signals through neural networks without altering the model.
02Testing whether explanations support understanding, appropriate trust, and decision quality.
03Building adaptive and physics-guided models for heterogeneous real-world time series.
04Aligning computational metrics with human perception and application-level evidence.
Introduces a non-interventional framework for tracing predictive information through neural networks and translating internal representations into human-centered evidence.
Investigates how explanation modalities influence trust, understanding, and decision making, placing human outcomes at the centre of explainable-AI evaluation.
Uses hypernetworks and learnable kernels to adapt energy forecasts across diverse consumer profiles.
Introduces a perceptual score that combines global and local evidence to evaluate the photorealistic quality of AI-generated images.
Combines learned temporal memory with physical guidance to improve the reliability of building-energy modeling.
Develops an order-theoretic tournament-graph method for assessing internal consistency in structured comparative settings.
Research statement“Accuracy is only the beginning. High-impact AI must also expose evidence, withstand rigorous testing, and make sense to the people who rely on it.”
I welcome collaborations in mechanistic interpretability, human-centered XAI, energy AI, and rigorous evaluation.
Start a conversation