Machine learning engineer specializing in wearable health AI — ECG, PPG, BCG, and ICG signals translated into algorithms running on devices in millions of hands and wrists.
I build ML and deep learning systems for physiological signals — the kind that have to work reliably on-device, in real time, under real-world noise. My background spans Samsung R&D, where I shipped commercialized algorithms to the Galaxy Watch, and HealthTech startups building next-generation cardiac and sleep monitoring. Along the way I've earned a patent in glycemic anomaly detection and contributed to peer-reviewed research. My core strength is the full path: from raw sensor stream to validated, deployed model.
Contactless heart-rate and HRV estimation from ballistocardiogram signals on a bed-based system.
95% accurate with Polar referenceSleep-stage classification from BCG signals, enabling unobtrusive overnight monitoring without wearables.
DeployedMultimodal snore detection fusing audio and BCG from a bed-based sensing system.
MultimodalHeart-rate and HRV pipelines built on PPG and ECG data for multiple HealthTech startups.
Multi-clientBlood-glucose and blood-pressure prediction models shipped for the Samsung Galaxy Watch.
Shipped to productReal-time cycling detection from accelerometer and gyroscope data, running on-device for millions of users.
Shipped to productActivity-recognition models for Galaxy Watches and a range of startup wearables across motion-sensor data.
Cross-deviceShipped cardiac output, stroke volume, PEP, and ejection-time estimation from impedance cardiography data.
ShippedA method to detect and analyze anomalous glycemic fluctuations in users, strengthening predictive healthcare analytics.
Organization: SamsungPeer-reviewed research applying ML to chemical reactivity prediction.
Saini, Sharma & Nivatia (2022)