sriram Yalamati

About

ML/NLP practitioner with healthcare payer experience at HCSC, focused on connecting messy unstructured clinical notes to structured claims/provider data to improve fraud-analytics workflows. Has hands-on experience fine-tuning transformers in AWS SageMaker, building large-scale embedding search with FAISS, and implementing robust entity resolution using golden datasets, precision/recall calibration, and production monitoring for drift.

Experience

Education

Key Strengths

• Applied clinical NLP to link unstructured clinical documents with structured claims data for fraud analytics
• Fine-tuned transformer models with domain-specific tokens using a labeled dataset
• Rigorous model evaluation (precision/recall/F1, cross-validation, testing on unseen notes, clinical team manual validation)
• Strong entity resolution approach combining multiple similarity algorithms and secondary features for missing identifiers
• Uses golden datasets and threshold calibration to control false-merge rates
• Built embedding-based similarity search at scale (bi-encoder + FAISS) and measured impact via top-k accuracy
• Production-minded workflow design with monitoring for confidence/drift and modular, testable pipeline structure