I build the validation layer underneath agentic AI — retrieval pipelines, agent reasoning chains, and evaluation infrastructure that decides whether a model's output can be trusted.
I work at the point where a model's output has to be checked before anyone can trust it — retrieval pipelines, agent reasoning chains, and the validation logic that scores groundedness and confidence before escalating what a system isn't sure about.
Alongside that, I build the backend infrastructure those systems run on: modular services, event-driven pipelines, and CI/CD that ships them reliably.
I hold two patents — one published, one filed — and I'm finishing an Integrated M.Tech in CSE with Business Analytics at VIT Chennai, graduating 2027.
Agentic clinical assistant combining metadata-aware retrieval, structured generation, and validation agents for groundedness, schema compliance, confidence scoring, and human escalation.
Event-driven backend for real-time vehicle tracking, route optimization, and maintenance alerts, streaming status updates through Kafka.
Policy-driven backend for asset lifecycle, approval workflows, inventory allocation, audit logging, and SLA escalations with strict role-based access control.
Parallelized ETL pipeline consolidating interconnected ERP datasets — inventory, vendors, invoices, deliveries — into analysis-ready outputs via CLI-driven batch workflows.
GlitchCon '26 winning project — full-stack application built under hackathon conditions, shipping production-grade features in a compressed sprint cycle.
Diabetic retinopathy detection pipeline using medical image classification — trained to identify disease stages from fundus images using deep learning.
Open to Backend, Applied AI, and Frontier AI Engineering roles. Based in Chennai — remote-first, travel when needed.
cavinchandran@gmail.com ↗