Writing

Notes on responsible AI

On fairness in clinical machine learning, AI governance, and building LLM systems that can show their evidence.

Trustworthy AIHealthcareFairness

Why Accuracy Isn't Enough: The Case for Trustworthy Medical AI

A model can be 99% accurate and still fail the patients who need it most. How fairness, transparency, and reliability change what we build.

6 min read →
Fairness MLOpen SourceHealthcare

Mitigating Demographic Bias in Clinical Models: Lessons from M-TRUST

Reducing demographic bias by 30.8% without sacrificing performance — across demographic, annotation, and amplification axes.

8 min read →
LLMRAGClinical AI

Grounding LLMs in Evidence: Building Citation-First Clinical RAG

A ‘No Source, No Result’ approach to retrieval — thematic clustering, NLI fact-checking, and source-level attribution for traceable answers.

7 min read →
StoryImpactOgroPath

From Rural Bangladesh to Production AI: Democratising Healthcare

Why limited access to care shaped my research and led to founding OgroPath — and what accessibility-first AI really demands.

5 min read →