I'm a self-taught, self-funded founder based in Sri Lanka, finishing a BSc in Software Engineering at Birmingham City University remotely. I co-founded PartnerLabz, a profitable software company with international clients, and PathFinder, an AI product I've built and grown independently with no marketing budget or outside investment.

I'm currently starting a research project with my co-founder Chethiya De Silva (PathFinder's CTO) to benchmark why current AI detection models fail on species they've never seen before accuracy drops from 82% to 43% on novel species, and most systems force a wrong label instead of flagging real uncertainty. This connects to a documented, real bottleneck: over half of Plymouth Marine Laboratory's 9,000+ deep-sea imagery deployments remain unanalyzed for exactly this reason.

I care about building things that are used, not just published, this project's first output (a benchmark paper and open evaluation protocol) is deliberately structured to be small, falsifiable, and useful to other researchers, before we commit to building the full model.