Indoor air can be up to 5x more toxic than outdoor air

You can't see it.
You're breathing it right now.

Every day, invisible spikes in CO₂ and toxic particulates build up in your home and office that you'd never know. PoWear is a wearable safety device that senses hazardous air in real time and shows you exactly where to act, before it affects your family's health.

89.1M

Real-world readings behind our detection model

77%

Faster reduction of hazardous CO₂ levels

≤34 μs

Real-time on-device hazard alerts

100% privacy with zero Cameras or microphones

How PoWear Protects You

Sense the danger. See it. Stop it.

Three technologies work together in every PoWear to catch hazardous air before it catches you.

Step 01 — Sense

The Detection Engine

Trained on 89.1 million real-world readings from 30 homes, PoWear's sensors know exactly what a dangerous spike in CO₂ or particulate matter looks like, before it's harmful.

Step 02 — See It

The PoWear Wearable

The wristband that started it all. PoWear turns invisible pollution into a glowing AR bubble you can point a fan at, so you fix the danger zone in seconds, not hours.

Step 03 — Stay Safe

The On-Device AI

On-device AI learns your household's risky habits like frying, poor ventilation, overnight buildup and warns you before they happen again. No cameras, no cloud, no privacy risk.

Live · Healthy

CO₂

612 ppm

VOC

0.3 mg/m³

Particles

8 µg/m³

Heart Rate

68 bpm

SpO₂

98%

23°C 45%

Proven In Real Homes

PoWear doesn't just warn you. It works.

Tested across 6 real home and office setups, PoWear drove hazardous CO₂ levels down to safe levels dramatically faster than doing nothing and hoping for the best.

Indoor CO₂ Reduction

Target safety baseline threshold set to 800 PPM across all deployments.

BeforeAfter using PoWear

Early Access Program

Be the first to wear PoWear.

Join the waitlist to get early access, launch pricing, and updates as we roll out PoWear to homes and offices.

Join Waitlist

Not Just Another Gadget

Backed by peer-reviewed science.

PoWear isn't a hunch, it's built on 89.1 million real sensor readings and published, peer-reviewed research. The full dataset, firmware, and models are open for independent verification.

DALTON Dataset Repository PoWear ESP32 BLE Firmware (coming soon) PoHAR Foundation Models (comming soon)

citation_bibtex.bib

@article{karmakar2024indoor,
    title={Indoor air quality dataset with activities of daily living in low to middle-income communities},
    author={Karmakar, Prasenjit and Pradhan, Swadhin and Chakraborty, Sandip},
    journal={Advances in Neural Information Processing Systems},
    volume={37},
    pages={70076--70100},
    year={2024}
}

@inproceedings{karmakar2026invisible,
    title={From Invisible to Actionable: Augmented Reality Interactions with Indoor CO2},
    author={Karmakar, Prasenjit and Yadav, Manjeet and Rout, Swayanshu and Pradhan, Swadhin and Chakraborty, Sandip},
    booktitle={Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems},
    pages={1--20},
    year={2026}
}

@article{karmakar2026pohar,
    title={PoHAR: Understanding Hyperlocal Human Activities with Pollution Sensor Networks},
    author={Karmakar, Prasenjit and Reddy, Karthik and Chakraborty, Sandip},
    journal={arXiv preprint arXiv:2605.09434},
    year={2026}
}

The Detection Engine Behind PoWear

DALTON: How PoWear Knows Danger When It Sees It.

Most air monitors guess using a single sensor in one corner of a room. PoWear doesn't guess. Its detection model is trained on DALTON, 89.1 million readings from multi-sensor grids across 30 real homes over 13,646 continuous hours, so it recognizes exactly how everyday habits turn safe air dangerous.

Real Dangers We Found In Real Homes

The Deep-Frying Migration Hazard

Indian culinary habits frequently involve long-term deep frying with spices. This generates massive spikes in sub-micron particulate matter (PM₂.₅) and Volatile Organic Compounds (VOCs). Unlike boiling, which stays contained in the kitchen, frying aerosols aggressively migrate into bedrooms within 15 minutes.

The Overnight Ghost Effect

Unwashed dishes or food scraps left overnight in kitchen sinks cause a continuous, 8-hour rise in ambient VOC and Ethanol (C₂H₅OH) concentrations. Even with no active occupants, indoor air toxicity steadily degrades until the sink is sanitized.

The Architectural Trap

Floor plans with isolated kitchen structures trap VOCs aggressively. While CO₂ levels might quickly normalize via basic window ventilation, dangerous VOCs linger significantly longer, getting caught in architectural blind spots.

Sensor Grid Specifications

Microcontroller:ESP32 Dual-Core 240MHz

CO₂ Sensor:MH-Z16 NDIR Infrared

Particulate Sensor:Plantower PMS5003 Laser

Sampling Frequency:1 Hz (Synced via NTP)

Annotation Engine:Android Speech-to-Text API

The Wearable Safety Device

Meet PoWear v1.

Carbon Dioxide (CO₂) is odorless and silently pools in "dead zones" around desks and beds. You open a window and assume the room is fine, but the hazard is still sitting right next to you. PoWear is the wristband that catches it, shows you exactly where it is, and tells you how to clear it.

A Warning You Can Actually Act On

A generic "air quality bad" notification doesn't tell you what to do. That's why PoWear shows the hazard exactly where it's forming. The moment PoWear detects CO₂ exceeding 1200 PPM on your wrist, it pushes the danger zone straight to your phone or AR headset — so you know precisely where to point a fan.

Volumetric Color Coding: Blue spheres indicate baseline air (<800 PPM), green represents moderate accumulation (800–1200 PPM), and pulsing red spheres warn of hazardous stagnation (>1400 PPM).
Targeted Airflow Nudging: Guided by visual spheres, users do not just open windows blindly; they position desk fans to blow directly into the stagnant red bubble cluster, reducing clearance times from hours to seconds.

Tested With Real Users (N=32)

84.3%

Preferred PoWear's AR alerts over a plain 2D app

3.8x

Faster to clear a hazardous hotspot

Try It Yourself1650 PPM — Hazardous

Click to
Ventilate

Target desk zone hotspot

This is what you'd see on your phone: click the red danger zone to simulate pointing a fan at it, and watch how fast a targeted response clears the hazard, versus just hoping it goes away on its own.

The Smart AI Inside PoWear

PoHAR: Protection Without Surveillance.

A safety device that watches you through a camera isn't safety!! it's a new risk. PoHAR is the on-device AI that learns your household's risky patterns from air chemistry alone. Every time someone cooks, cleans, or leaves dishes overnight, PoWear recognizes the pattern and warns you, without ever recording you.

The Decentralized Edge Intelligence Pipeline

Because standard IoT microcontrollers have tiny memory footprints (4MB Flash), they cannot process raw high-frequency sensor streams centrally. PoHAR introduces three novel algorithmic primitives running natively on ESP32 hardware:

01. Time-Frequency Consistency (TF-C) Embeddings

On-device self-supervised learning models compress multivariate time-series air data (PM₂.₅, CO₂, VOC, Temp, RH) into compact, low-dimensional spatial vectors without cloud offloading.

02. Ultra-Lightweight RAFT Consensus

Distributed sensor nodes continuously elect a local "Cluster Leader" using heartbeat sync timeouts. Only the leader executes the final activity classification, reducing household energy draw by 68%.

03. Conflict-Free Replicated Data Types (Set-CvRDT)

Ensures seamless data merging across nodes even if household Wi-Fi drops or packets are delayed, maintaining 97.41% general daily activity accuracy.

RAFT Leader Election Simulator

ESP32-01
Follower

ESP32-02
Follower

ESP32-03
Follower

ESP32-04
Follower

ESP32-05
Follower

ESP32-06
Follower

Algorithm: Set-CvRDT + SSL EmbeddingsInference Latency: 0.00 μs

Cooking Recognition Precision

99.68%

Differentiates frying, boiling, and sweeping via chemical diffusion gradients alone without optical cameras.

Sensing Modality	Privacy Risk	Culinary Precision	Compute Overhead
Optical Video Cameras	Extreme (Visual)	98.20%	High (GPU/Cloud)
Acoustic Microphones	High (Audio)	91.40%	Medium (DSP Required)
PoHAR Air Network	0.0% (Zero-Risk)	99.68%	≤ 34 μs Latency

See It Work

Don't just take our word for it.

Watch PoWear catch a real hazard, guide a real fix, and clear the air — filmed in real homes and labs, not a studio.

DALTON Archive

13,646 Hour Indoor Field Study & Methodology

DALTON — Study Teaser & Methodology 4K

Part 1: Spatiotemporal Frying Particulate MigrationYouTube

DALTON — Deployment & Sensor Telemetry 1080p HD

Part 2: Multi-Room Sensor Hardware ArchitectureYouTube

PoWear Archive

Augmented Reality Spatial Air Bubbles

PoWear — Real-Time AR Bubble Nudging & Ventilation Fan Actuation 1080p HD

Research Highlight: Visualizing localized 1Hz CO₂ telemetry in spatial headsets to guide targeted human airflow intervention.

Open in YouTube

PoHAR Archive

Zero-Video Edge ML Activity Recognition

PoHAR — Self-Supervised Edge Inference & RAFT Consensus Demo 4K

Research Highlight: Demonstrating ≤ 34 μs latency classification of culinary activities without optical cameras via chemical diffusion gradients.

Open in YouTube

You can't see it.
You're breathing it right now.