Reading Risk Early: From Asthma Exacerbation to Financial Markets

ABSTRACT

Roughly 25 million Americans have asthma, and about ten die each day from severe exacerbations as rescue medication often arrives after the attack has already happened. Consumer wearables are descriptive rather than predictive as heart rate, blood oxygen, and activity are collected as independent streams, and rarely processed on devices leaving the burden of care on a patient for detecting an attack.

This talk frames exacerbation onset as a forecastable state transition and presents the wearable architecture built around it, the core technology behind my startup BioProxal. Photoplethysmography and accelerometry feed quantized machine-learning sub-agents of baseline modeling, anomaly detection, and risk assessment that are coordinated by a deterministic finite-state orchestrator running entirely on-device with no raw signal leaving the wearable. Quantization and sensing policy cannot be designed independently, since a single fixed model footprint cannot stay dynamic and personalized to each patient. I will cover the architecture and patent, cold-start personalization, and early results in detecting attacks.

The second half focuses on how this technology generalizes as a risk signal problem. The math behind detecting attacks with AI maintains the same fundamentals in early-stage drug discovery and in detecting risk at large financial firms.

BIOGRAPHY

Michael Moschello is a student at Stevens Institute of Technology, concurrently pursuing his BEng and MEng in Computer Engineering and Applied Artificial Intelligence. He is also the founder of BioProxal, a wearable platform that predicts severe asthma episodes before they occur. He is the inventor on a patent covering on-device agentic inference for physiological exacerbation-risk prediction using quantized machine-learning sub-agents under bounded memory and power budgets.

Moschello developed BioProxal through the Stevens Launchpad program, winning top prize for his presentation at BioNJ's AI & Digital Health Day and appearing alongside Stevens President Nariman Farvardin at Prudential's Newark headquarters for a Data, Service and Technology Town Hall. He spent two summers at Merck & Co. in various AI/ML research and development roles and is an incoming analyst at BlackRock.