•Innovated machine learning solution to predict ovulation from voice, achieving 82% accuracy and enabling targeted family planning.•Drove creation of ML models detecting hormonal fluctuations in female voices with 89% accuracy in identifying fertile window.•Led productionization of models via Flask APIs and containerization with Docker. Architected GitHub Actions-based CI/CD pipeline deployments onto AWS Elastic Beanstalk, enabling robust, scalable, and reliable access to these machine learning solutions.•Pioneered non-invasive ML models predicting blood pressure and blood sugar from voice data with RMSE of 5 and 15. Reduced deployment costs by 40% by architecting a serverless solution of AWS Lambda functions triggered by AWS DynamoDB updates for efficient delivery

•Developed ML model to detect word boundaries in continuous Marathi speech (91% accuracy), enabling improved language processing.•Analysed audio features using statistical methods, identifying vocal biomarkers with correlation to medical conditions.•Built a classifier to extract voiced regions from audio data (precision: 0.89, recall: 0.86), improving the performance of downstream models