My work spans agentic AI, large language models and retrieval, computer vision, medical AI, and MLOps — the full path from a research idea to a deployed, observable service.

Right now that means multi-agent platforms with real planning and governance, safe tooling that lets an agent touch production systems without a security team losing sleep, and reliability work on medical-imaging models that quietly degrade after deployment. My MSc dissertation is on detecting and explaining that drift.

What ties it together is a bias toward engineering discipline: honest baselines before novel methods, an evaluation harness for every project, calibrated confidence over confident guessing, and metrics that are measured rather than estimated. I publish negative results too — a failed localisation head is still a finding.

Before the MSc I completed a B.Tech in Computer Science & Engineering in India, specialising in AI and machine learning.