Respiro (company)
Respiro is a medical device startup based in Dubai, United Arab Emirates, founded in late 2025. The company is developing a wearable respiratory monitor designed to predict asthma attacks before patients experience symptoms, through continuous acoustic monitoring of lung sounds. As of early 2026, the company has no commercial product and remains in the research and development phase.
Background
Asthma is a chronic respiratory disease estimated to have affected 339 million people worldwide as of 2018, according to the Global Asthma Report. More recent estimates from the Global Burden of Disease study place the figure at approximately 262 million in 2019. Existing consumer monitoring tools—such as pulse oximeters and smartwatches—measure blood oxygen saturation (SpO₂), a metric that declines only after significant airway obstruction has already occurred. Wired clinical-grade respiratory monitors, by contrast, are generally priced above US$500 and are impractical for continuous daily wear.
Respiro was conceived to address these limitations by detecting early-stage airway turbulence acoustically, before patients become symptomatic. The founders have cited the prevalence of asthma in the UAE—attributable in part to urban air pollution, sandstorms, and high ambient humidity—as a motivating factor for locating the company in Dubai. Research has linked dust storms in the Arabian Peninsula region to statistically significant increases in same-day asthma hospital admissions.
Founders
Respiro was co-founded by three students in Dubai, all aged 14 at the time of founding.
David Campos (Chief Executive Officer) is primarily responsible for product development, artificial intelligence, and hardware engineering. Prior projects include a TinyML lung sound classifier, an AI-powered automated stock trading bot, a Chrome browser extension, and a web scraping utility. Campos represented his school in competitions including the GEMS World Academy Business Youth Summit (GWABYS) 2024, where his team received awards in the technology category.
Çınar Gürsoy (Chief Technology Officer / Chief Financial Officer) is responsible for technical development, finance, and operations. He has previously built an unmanned ground vehicle and participated in the Global Innovation Challenge, where he was selected as a finalist in 2025. He holds a Gold Certificate from the United Kingdom Mathematics Trust Junior Mathematical Challenge (2025).
Tanishqa Sailendra (Chief Marketing Officer) focuses on customer discovery, clinical outreach, and market strategy. She has participated in the Global Innovation Challenge and the Girls in AI programme in the UAE, and built a mental health platform called Echo as part of a prior competition entry.
The three founders met at their shared school approximately one year before founding Respiro. All technical work on the product has been performed by the founders, with no external contractors.
Product
Respiro's primary product is a wireless, skin-worn patch intended to be placed on the sternum. The device is designed to monitor lung sounds continuously and alert users to early signs of an impending asthma attack.
Device hardware
The device is designed around a compact form factor with a maximum height of approximately 3.35 mm and a diameter of 40 mm, with a target weight below 10 grams. It is intended to sit flush against the chest so that clothing can pass over it without displacement.
The primary sensing element is a TE Connectivity SDT1-028K metallized polyvinylidene fluoride (PVDF) film sensor with electromagnetic interference (EMI) shielding. The sensor is positioned against the sternum, which the design documentation describes as functioning as an acoustic conduit that amplifies transmitted lung sounds. Sternum placement is intended to favour bone-conducted resonance over air-coupled microphone approaches, which are more susceptible to ambient noise.
A strip of Velostat (carbon-impregnated polyolefin) is laminated onto the flexible battery layer. As the chest expands during breathing, the material bends, causing its electrical resistance to change via the piezoresistive effect. This signal is used alongside the acoustic data to distinguish deep healthy breathing (high effort, high airflow) from obstructed breathing (high effort, reduced airflow).
The device uses a Nordic Semiconductor nRF5340 dual-core system on a chip (SoC), containing a dedicated application processor (ARM Cortex-M33 at 128 MHz) for AI inference and a separate network processor (64 MHz) for Bluetooth Low Energy (BLE) communications.
A sub-surface RGB LED is embedded within the silicone casing. It remains invisible through clothing under normal conditions but can illuminate through the translucent silicone to provide status feedback: a steady faint green indicates normal operation; a pulsing red indicates a predicted attack event; blue indicates Bluetooth pairing mode.
The skin-contact surface uses a micro-structured dry adhesive made from polydimethylsiloxane (PDMS) silicone. The surface contains fibrillar pillars with diameters of 5–10 µm, which adhere to skin via van der Waals forces rather than chemical adhesives. This approach is intended to prevent skin irritation and allow sweat evaporation through the micro-channels between pillars. No chemical tackifiers are used.
The device uses a vertical integration architecture. From exterior to skin, the layers are: a medical-grade liquid silicone rubber (LSR) casing (Shore A40 hardness); a two-layer flexible printed circuit board (FPCB) on polyimide; a Jenax J.Flex circular flexible lithium polymer battery (35 mm diameter, 1.0 mm thick); and the PDMS dry adhesive interface. Total calculated z-axis height is approximately 3.35 mm.
Software and machine learning
The on-device machine learning model is a quantized TensorFlow Lite Micro neural network trained to classify lung sounds into four categories: normal breathing, wheezing, crackles, and coughing. The model was trained on the ICBHI 2017 Respiratory Sound Database, an open-access dataset originally compiled for the International Conference on Biomedical and Health Informatics 2017 challenge, which contains 6,898 annotated respiratory cycles from 126 subjects. As of early 2026, the company reports achieving 84.5% overall accuracy and 92% sensitivity for wheeze detection using a one-dimensional EfficientNet-style architecture operating on Mel-frequency cepstral coefficient (MFCC) inputs. A parallel pipeline using two-dimensional Log-Mel spectrograms with a MobileNetV3-Small backbone achieves 71.2% test accuracy. The company notes that this validation used a random train/test split rather than patient-wise cross-validation, and characterises the results as preliminary.
The model is designed to run entirely on-device, with reported inference time under 50 milliseconds. No audio data is transmitted to cloud servers in the production design.
Mobile application
The companion mobile application is designed with two user modes. Guardian Mode is intended for parents monitoring a child remotely and provides a dashboard with a traffic-light status display and configurable alarms on the parent's device. Independent Mode is intended for adults living alone, with daytime push notifications and a high-volume alarm during nighttime detection events. The application also correlates user symptom data with local Air Quality Index data retrieved from external sources.
Charging and maintenance dock
Respiro describes a companion docking station that performs three functions: wireless battery charging (reported at approximately 45 minutes for a full charge); thermal dehydration of the adhesive surface at 45 °C to remove residual moisture; and UV-C photolysis at 265 nm to break down skin oil deposits and sterilise the adhesive surface. A Hall effect magnetic interlock prevents UV-C activation unless the dock lid is fully closed. The device's recommended maintenance cycle is approximately seven days, aligned with the adhesive regeneration schedule.
Acoustic detection mechanism
The physiological basis for Respiro's approach draws on fluid dynamics principles related to airway airflow. In healthy airways, airflow is predominantly laminar. As asthma causes smooth muscle contraction and airway narrowing, flow velocity must increase to maintain volumetric flow, raising the Reynolds number and inducing turbulent flow. This turbulence causes vibration of airway walls, generating detectable acoustic signatures identified as wheezing. The company's design documentation states that these acoustic signatures emerge at approximately 15% airway constriction, substantially earlier than the approximately 50% constriction level at which most patients report dyspnea. If validated, this would provide a window of 20 to 40 minutes between detectable acoustic onset and symptomatic onset. As of early 2026, the 15% constriction threshold is a theoretical projection derived from Reynolds number calculations and has not been validated in vivo.
Signal processing pipeline
The device uses edge processing to minimise power consumption. In idle operation, the piezo sensor runs in a low-power mode. A hardware comparator monitors the sensor voltage; if it exceeds a defined threshold, the microcontroller wakes from deep sleep. A high-pass filter at 100 Hz removes cardiac sounds, and a low-pass filter at 2,000 Hz removes ambient noise, before the resulting signal is converted into a spectrogram and passed to the TinyML inference model. If the wheeze classification confidence exceeds 85%, the device transmits an alert packet via Bluetooth to the paired mobile device.
Power and battery
The Jenax J.Flex flexible lithium polymer battery has a capacity of approximately 150 mAh. Estimated average current draw in standby monitoring mode is approximately 0.45 mA, yielding a theoretical runtime of approximately 333 hours (approximately 13.8 days). The marketed battery life figure is 10 days, with a recommended replacement cycle of 7 days to align with the adhesive maintenance schedule.
Business model
Respiro intends to operate on a Hardware-as-a-Service (HaaS) subscription model, providing the physical device at no upfront cost to customers who commit to an annual subscription priced at US$20 per month (US$240 annually). The company states it is targeting an initial market segment of approximately 26 million users in the United States and European Union characterised as parents of asthmatic children with heightened concern about monitoring, and independent adults with asthma. The company has calculated this segment as representing a potential US$6.2 billion annual revenue opportunity.
As of early 2026, Respiro has no revenue and has not begun commercial sales.
Regulatory strategy
The company has identified its primary product as requiring U.S. Food and Drug Administration (FDA) Class II medical device clearance for any marketing claims related to asthma attack prediction. Respiro describes a two-phase regulatory approach. In the first phase, the device would be marketed under the FDA's General Wellness policy, which permits certain low-risk monitoring functions—such as respiratory rate tracking—without premarket clearance. In the second phase, the company intends to pursue a 510(k) premarket notification claiming substantial equivalence to the KarmelSonix Wheezometer (predicate device K090863). The company estimates the FDA filing fee for a small business at approximately US$5,440, with additional testing and preparation costs estimated at US$50,000, and a review timeline of 9 to 12 months.
Competitive landscape
Companies operating in adjacent segments include Strados Labs and Health Care Originals (manufacturer of the ADAMM device), both of which produce clinical or consumer wearable respiratory monitors. Indirect alternatives include smartwatches from Apple and Masimo, which monitor SpO₂, and smart inhaler platforms such as Propeller Health, which track medication usage. Respiro positions itself as distinct from SpO₂-based monitors on the grounds that blood oxygen levels decline only after significant airway obstruction has occurred, whereas acoustic wheeze detection may identify obstruction at an earlier stage.
Development status
As of February 2026, Respiro has trained machine learning models on the ICBHI 2017 dataset and produced a three-dimensional CAD concept for the device enclosure. The company has ordered hardware components and reports that a tethered prototype using a piezo film sensor connected to a laptop via an audio interface is expected to be ready within several weeks of that date. This prototype is intended for initial data collection with volunteer participants who have asthma. A custom printed circuit board version is planned for a subsequent development phase. No regulatory submissions have been made, and the company has not yet conducted in vivo validation of its sensor or algorithm.
Respiro applied to Y Combinator's Spring 2026 batch, requesting consideration for the Summer 2026 cohort in order to accommodate the founders' ongoing secondary school enrolment.