Understand before trusting.
Explainability should reveal evidence, assumptions, and failure modes—not just produce a prettier explanation.
Research → Postdoctoral Fellow · Applied AI
Muhammad Umair Danish, PhD
I build AI systems that people can understand, question, and trust.
My research connects explainable AI, machine learning, energy systems, time-series forecasting, and human-centered evaluation. I care about turning technical ideas into methods and software that remain useful beyond a paper.
LIVE research metrics
LIVE · syncing Google Scholar…
Research in numbers.
The person behind the work
Research is most useful when it stays connected to people.
I am an AI researcher who enjoys the full path from a difficult question to an experiment, a reproducible result, and something another person can actually try. My work is technical, but the goal is practical: make intelligent systems more understandable, reliable, and useful.
My background and journeyWork on real systems.
I focus on problems where models meet physical processes, people, operational constraints, and imperfect data.
Research →Leave something usable.
Where possible, I turn research into open software, reproducible examples, and tools others can inspect themselves.
Software →Selected research
View all publications →Three projects that show how I think.
Representative work spanning energy intelligence, perceptual evaluation, and human-centered AI.
Kolmogorov-Arnold Recurrent Network for Short-Term Load Forecasting Across Diverse Consumers
Contribution: Develops a recurrent architecture based on Kolmogorov-Arnold representations for forecasting heterogeneous consumer loads.
Why it matters: forecasting methods need to generalize across very different consumers—not only one convenient load profile.
Global-Local Image Perceptual Score (GLIPS): Evaluating Photorealistic Quality of AI-Generated Images
Contribution: Introduces a perceptual score that combines global and local evidence to evaluate photorealistic quality.
Why it matters: image-quality metrics should better reflect what people actually perceive.
Towards a Unified Evaluation Framework: Integrating Human Perception and Metrics for AI-Generated Images
Contribution: Integrates human judgments and automated image-quality metrics into a unified framework.
Why it matters: automated scores become more meaningful when interpreted together with human judgments.
Research areas
View all areas →Four connected directions.
Each area addresses a distinct part of building machine-learning systems that can be understood, evaluated, and applied responsibly.
01Explainable & Interpretable AI
Mechanistic and model-level methods for understanding learned representations and predictions.
Explore this area →02Physics-Guided Machine Learning
Learning systems that incorporate physical structure, constraints, memory, and domain knowledge.
Explore this area →03Time-Series & Energy Intelligence
Forecasting, robust temporal modeling, and adaptive learning across heterogeneous energy systems.
Explore this area →04Responsible & Applied AI
Human-centered evaluation and reliable machine learning across perception, imaging, finance, and applied domains.
Explore this area →Research software
Explore all software →Software you can actually open and use.
Open scientific tools that turn research ideas into inspectable, reproducible workflows.
InterpretabilityLIVE
Linear Lens
Non-interventional, human-centered mechanistic interpretability for neuron roles, layer fingerprints, and semantic pathways.
PreprocessingLIVE
Core-Norm
Bounded, asymmetric and invertible preprocessing for numeric machine-learning features with public benchmark evidence.
EvaluationLIVE
OpenMetricLab
Local-first scientific evaluation for regression, classification, segmentation, and image registration.
AlgorithmsLIVE
AlgoLens
Analyze source code for time and space complexity using a deterministic, explainable reasoning trace.
Recent milestones
All publications →What changed recently.
Accepted work is labeled as accepted rather than published.
2026
Accepted · Conference
Open record →
2026Cognitive Effects of Textual, Graphical, and Interactive Explanations in Explainable Artificial Intelligence
52nd Annual Conference of the IEEE Industrial Electronics Society (IECON 2026)
Accepted · Conference
Open record →
2026Human-Centered Explainable Artificial Intelligence and the Effects of Explanation Modalities on Trust, Understanding, and Decision Making
52nd Annual Conference of the IEEE Industrial Electronics Society (IECON 2026)
Accepted · Journal
Open record →
Jun 9Dispersity Measures Within Sessions During Improvised Active Music Therapy with Clients with Parkinson’s Disease
Approaches: An Interdisciplinary Journal of Music Therapy
Academic milestone · 2026
View timeline →
PhD Convocation at Western University
PhD in Electrical and Computer Engineering · Western University
—Today
Daily Knowledge · On this day
— Loading today’s historical note…
A short historical event is loaded from Wikipedia’s “On this day” feed.
Read on Wikipedia ↗Source: Wikipedia · CC BY-SA 4.0
Collaborators & team
Research is a team activity.
People I work with across explainable AI, energy intelligence, human factors, and interdisciplinary research.
Collaboration
Good research gets better through conversation.
I am open to research collaborations, applied AI projects, academic opportunities, and technical discussions around interpretable machine learning and research software.
LIVE GLOBAL REACHVisitors from around the world.
A live geographic view of visitors reaching this website.
Live map blocked in this browser.Open live visitor statistics →