At Innovaccer, I architected FlowCode, a production clinical coding system that orchestrates 15 agents and achieved 94% precision extracting ICD-10-CM, CPT, and HCPCS codes. I also designed safeguards for hallucinations and fault-tolerant recovery.
I architected Incurate, a clinical NLP platform that processes medical documents and maps extracted entities to codes. Fine-tuning clinical language models raised entity extraction F1 from 65% to 94%, while migrating from AWS Comprehend Medical to open-source infrastructure cut annual platform costs from $90K to $5K.
Earlier, at AIMonk Labs, I developed an SVM-based customer segmentation model and deployed real-time anomaly detection on in-store camera feeds. I also built a project that uses multiple LLMs to create validated medical knowledge graphs.
