At PocketHealth, I build healthcare conversational AI that connects EHR data, clinical guidelines, and treatment protocols through RAG and HL7/FHIR standards.
I’ve engineered Neo4j knowledge graph pipelines, multi-agent systems using LangGraph and CrewAI, and fine-tuned GPT and LLaMA models on proprietary clinical data. I also established LLMOps practices with MLflow, Weights & Biases, and human-in-the-loop review to improve reliability, traceability, and compliant AI lifecycle management.
Previously at BenchSci, I developed NLP and semantic enrichment models for large-scale unstructured content, using entity linking, embeddings, topic modeling, clustering, sentiment analysis, and A/B experimentation.
At Deep Genomics, I led applied research in anomaly detection and unsupervised learning, turning Python research prototypes into production-grade ML pipelines and contributing research papers, patents, and internal whitepapers.
