Senior Engineering Manager, AI/ML · Qoria
Adjunct Senior Lecturer · Murdoch University
I build production ML systems and research how to make them explain themselves — bridging computational pragmatics, NLP, and applied engineering.
I'm an AI engineer with a PhD in AI from Murdoch University, where my research centred on computational pragmatics and natural language understanding how meaning is inferred beyond literal text. That foundation now shapes how I think about explainability: not just what a model predicts, but why, and how to communicate that clearly to the people who rely on it.
Day to day, I lead AI/ML engineering at Qoria, and I conduct research with my team of research students as an Adjunct Senior Lecturer at Murdoch University. Alongside that, I'm building dakma — an SDK for ML explainability and compliance, named for the Sinhala word for vision and visibility.
Three threads that connect the research and the engineering.
SHAP-based methods, counterfactual explanations, and monitoring for explanation drift in production systems.
Computational models of inferred meaning — how systems can understand more than the literal text in front of them.
Shipping ML pipelines that hold up at scale — instrumentation, MLOps, and engineering teams that build them.
Leading AI/ML engineering, from research prototypes through to production systems.
Teaching and supervising PhD and Master's Research students in AI, continuing the research line from my PhD.
Led the AI function, setting technical direction and building out the team's ML capability.
Led data science work, building ML-driven solutions as part of the engineering team.
dakma-sdk reaches 2.0: Every Forecast Gets an Explanation
Read more →dakma-sdk reaches 1.0 with a stable decision event and explain_text, which attributes text decisions to individual words.
Read more →An introduction to an SDK on explainability and transparency.
Read more →Starting this blog to think in public about explainable AI, pragmatics, and what it actually takes to ship ML systems people can trust.
Read more →