The Center for Data Innovation recently spoke with Lisa Haxel, co-founder of Alveos, a New York-based company developing a wearable device that measures breathing directly from the chest. Haxel explained how the device detects small chest movements, uses an AI system to interpret breathing patterns, and converts those signals into clear, real-time physiological data.
David Kertai: What does Alveos offer?
Lisa Haxel: Alveos offers a small magnetic wearable that mounts on your chest, directly measures breathing, and delivers real‑time physiological feedback. Breathing is tightly linked to autonomic regulation—the body’s automatic control of functions like heart rate, stress responses, and recovery—yet it’s rarely measured accurately in daily life. Most consumer devices only estimate it indirectly through heart rate or body movement, and traditional chest‑based tools have been too bulky for everyday use.
To address this gap, we designed a device that captures a mechano‑acoustic signal by measuring the tiny vibrations in the chest wall created by each inhale and exhale. This data feeds into an AI system within the companion app, which interprets breathing patterns, detects shifts in physiological state, and provides short, targeted breathing guidance while verifying whether those interventions are effective.
Kertai: How does the device detect users’ breathing?
Haxel: The system uses a contact‑based sensor paired with motion sensing to pick up tiny vibrations from the chest wall, which naturally blocks out most surrounding environmental noise. After capturing this raw signal, we run it through a series of processing steps that clean and interpret it. This includes signal processing, which removes unwanted noise, and model‑based filtering, which uses learned patterns of real breathing to separate true physiological signals from everything else.
The system looks for breathing‑related physiological patterns, the rhythmic rise and fall of the chest, the timing of inhales and exhales, and the subtle mechanical signatures that accompany each breath. At the same time, it suppresses motion artifacts—vibrations caused by movement, like walking or shifting posture—and non‑respiratory signals such as speech or other body sounds. Finally, the system checks that the signal follows breathing patterns and only analyzes those segments, ensuring that the AI system works with accurate, high-quality data.
Kertai: What does your AI system look for in users’ breathing, and what type of feedback does it provide?
Haxel: Our AI system looks at how someone breathes, not just how fast. It analyzes the timing between inhales and exhales, how regular the pattern is, and estimates how deep each breath is. These patterns reflect a person’s autonomic state, meaning how their body is responding to stress, rest, or recovery.
To make this meaningful, the system compares each user to their own normal patterns rather than to population averages. It continuously updates its internal models as it observes more data, allowing the guidance it provides to become increasingly personalized.
The feedback itself is immediate and context‑aware. Instead of showing dashboards, the system delivers short, targeted breathing prompts, through haptics through the wearable device or simple cues on the app. For example, it might suggest to slow your breathing when you’re starting to feel stressed, or guide you through a few steady breaths to help your body recover after physical effort. Because the device measures the physiological response in real time, users can see whether the intervention worked. Over time, this also allows the AI system to learn which breathing strategies are most effective for the user.
Kertai: What challenges have you faced while developing Alveos?
Haxel: Ensuring reliability in real-world conditions has been one of our biggest challenges. Breathing signals can vary based on movement, device placement, and individual differences, so we need the system to perform consistently across many situations. We address this by tightly integrating the hardware, signal processing, and machine‑learning components so the system can adjust when real‑life conditions change, such as when someone starts walking, shifts their posture, or places the device slightly differently on their chest
Another challenge is translating complex physiological data into simple, intuitive guidance. We focus on turning subtle changes in breathing into clear prompts, such as slowing the exhale. At the same time, we design our AI system to avoid reacting too strongly to short-term disruptions, like a cough or sudden movement, so the feedback remains reliable.
Kertai: Are there any new features or innovations you’re excited to share?
Haxel: We are especially excited about deeper personalization over time. As the AI system collects more data, it becomes better at predicting when a breathing intervention will help and how to tailor it to each individual. We are also advancing other AI systems that combine breathing data with other contextual signals, such as movement or activity. This improves both accuracy and interpretability. Over time, this approach will enable more precise, real-time guidance for managing stress, improving recovery, and supporting performance in everyday situations.
