At Strand Life Sciences, I shipped TraitKB, an 8-node LangGraph pipeline for genomic evidence curation that cut first-pass work from analyst-days to about 10 minutes per trait. Its strand-aware variant annotation achieved 96–97% agreement with expert curation on an LLM-free harness.
I also own a production biomedical metadata pipeline that extracts and validates study fields against curator ground truth, and contributed detection and tissue-of-origin classifiers to the EMERGE multi-cancer early detection study. My work spans production LLM systems and applied ML, with research training from IISc.
