Mihira Gudimetla

At Humana, I build and deploy health risk stratification, fraud detection, drug adherence, and churn prediction models on AWS SageMaker. I own workflows from feature engineering and validation through production deployment and monitoring.

I’ve built NLP models with Hugging Face Transformers, BERT, SpaCy, and NLTK to extract clinical entities, intent, and sentiment from unstructured healthcare data. I also prototyped a LangChain and Claude API clinical decision-support agent for natural-language population-data queries.

Previously at Johnson & Johnson, I fine-tuned BERT models for physician notes and lab reports, developed hospital readmission risk models, and established MLflow-based experiment tracking and model versioning. I communicated model behavior and limitations to clinical, regulatory, and leadership stakeholders using SHAP-based risk interpretation.

My foundation at HSBC includes Python and SQL data pipelines, AWS Glue ETL, customer segmentation, credit-risk analytics, forecasting, and Power BI reporting. I work closely with data engineering, clinical, actuarial, compliance, and platform teams to turn complex data into dependable production outcomes.