At MPIRIQ, I built a clinical-trial matching engine that parses criteria with LLMs and scores matches, reducing manual screening effort by ~80%. I also migrated trial and cohort matching to a hybrid SQL + LLM architecture, reducing costs by ~2.7×.
I designed and shipped a cohort query platform that reduced cohort creation from hours to minutes, and parallelized an LLM extraction pipeline that increased throughput by ~5×. I also built an event-driven notification system on GCP using Terraform and Kubernetes.
Previously, at Gravitas AI, I improved an Indian Council of Medical Research chatbot’s response accuracy from 65% to 97%. At TechEnhance IT Services Pvt Ltd, I built legal-document retrieval and invoice-automation systems, and fine-tuned open-source LLMs including Llama, Falcon, Gemma, and Mixtral.
