Interpretable AI
Models that expose meaningful structure through latent traversals, reconstructions, and counterfactual reasoning.
About
Researcher, builder, and technical storyteller.
I’m a machine learning researcher working at the intersection of interpretable AI, scientific computing, medical imaging, and climate data.
Short bio
I build models that help us understand hidden structure.
I am a PhD candidate in Machine Learning at Federation University Australia, working within the Institute of Innovation, Science & Sustainability. My research focuses on interpretable representation learning using generative-discriminative models for ophthalmic imaging and tropical meteorology.
The common thread across my work is the idea that machine learning should do more than produce accurate predictions. In scientific settings, useful models should help researchers inspect transitions, explore counterfactuals, and understand why different cases behave differently.
My current research develops a Mixture-of-Encoders β-VAE framework for learning traversable latent spaces across heterogeneous data, including OCT-derived retinal thickness maps and spatio-temporal atmospheric tensors.
What I work on
Connecting machine learning with scientific questions.
Medical Imaging
Representation learning for OCT-derived RNFL and GCIPL maps to study glaucoma-related structural change.
Climate Data
Spatio-temporal modelling of tropical cyclone genesis using atmospheric variables, OWZ events, and ERA5 fields.
Research philosophy
I’m interested in models that do more than classify.
Prediction is useful.
Classification, AUC, and accuracy matter — especially when models are used as measurement tools or decision support systems.
But explanation is the scientific value.
I care about what the model learns, which latent directions are meaningful, and whether generated changes are clinically or physically plausible.
The goal is exploration.
A good model should let us ask: what would need to change for this case to look healthier, riskier, stronger, weaker, developing, or non-developing?
Journey
A path through computing, research, and scientific AI.
Nepal
Early foundations in curiosity, technology, and problem solving.
Australia
Academic development and postgraduate research training.
Machine Learning
Generative models, representation learning, and scientific computing.
PhD Research
Interpretable AI for retinal imaging and tropical cyclone genesis.
Next
Want the technical version?
Explore the research pages for the method, scientific domains, model architecture, and placeholder spaces for future results.
Explore research