Apr 2026 — Present
- –Architected multimodal wearable data pipeline unifying NHANES accelerometry, MMASH beat-to-beat HR, and Fitbit telemetry into a unified feature store; enabling circadian, sleep, and HRV signal extraction at scale.
- –Engineered biological age prediction system combining XGBoost with SHAP-driven explainability, surfacing wearable-derived aging signals previously inaccessible without clinical testing.
- –Fine-tuning Llama 3.2 1B via QLoRA on structured wearable serializations to generate interpretable biological age insights; quantizing to GGUF for CPU-deployable inference without GPU dependency.