P-01
Behavioral anomalies go unnoticed
Pets can't tell you when something's wrong. Owners usually catch problems only after symptoms are obvious — early signals like reduced activity or disrupted sleep slip by, and the intervention window closes.
Shipping solution · North America01 / Pet CollarIoT × AI Solution
Every heartbeat,
gently sensed
A smart pet collar built for the North American market. Fusing a 6-axis IMU, temperature sensing, and GPS — it gives pet owners light health monitoring plus real-time tracking, in one. Not medical diagnostics; just everyday care that reads what your pet's day really looks like.
nRF52840 BLE 5.06-axis IMUGPS / LTE-MIP67 waterproof14–21 day battery
6-axis
IMU sensing
Activity · posture · sleep
21d
Battery life
BLE + weekly + sparse GPS
IP67
Rated
All-weather indoor / outdoor
<10s
Fix latency
GPS · LTE-M handoff
Companion · livev1.4
Mochi · 4y · Shiba
Steps today 8,420 · Active 1h 42m
- Activity index84Active
- GeofenceAt home12m ago
- Skin temp38.4°CNormal
- Collar76%~13 days
BLE · GPS · LTE-M fusion · desensitized sample
§ 01 / The Problem
Four real pain points in pet health monitoring
The global pet wearables market grows 16%+ annually, but existing products still fall short on health monitoring accuracy, battery life, waterproofing, and breadth of use cases. This solution starts from real user pain points and redefines what light health + tracking should feel like together.
P-02
No continuous temperature tracking
Traditional temperature taking is manual — no 24-hour continuous monitoring. Fever or hypothermia is usually discovered at the vet's office, with no historical data to support diagnosis.
P-03
Lost pets and location blind spots
Millions of pets go missing every year. Traditional tags and chips only identify pets after they're found. Owners need to know location in real time — especially outdoors and while traveling.
P-04
Compromises in existing solutions
Existing products either offer GPS-only with no health monitoring, or use optical PPG heart rate sensors that fail (<60% accuracy) on long-haired breeds. Solutions that genuinely fuse light health + tracking remain rare.
§ 02 / The Approach
Three paths evaluated, one chosen
We systematically evaluated the three mainstream technical paths across four dimensions — breed compatibility, battery life, health monitoring depth, and competitive differentiation. The IMU + temperature + GPS fusion path won as our core tech stack.
A
GPS-only tracking
Tractive / Fi-style
High positioning accuracy, mature tech
Rich power-optimization recipes
Strong user awareness, market-validated
No health monitoring capability
Heavy feature homogenization
No differentiation moat
Health monitoring
None
Battery
7–30 days
Differentiation
Low
Tracking only — doesn't address health needs
B
PPG optical heart rate
PetPace / Invoxia-style
Continuous heart rate data
Clinical-grade health monitoring potential
Rich data dimensions
Very low accuracy (<60%) on long-haired breeds
High power draw, 3–5 day battery
Strict sensor contact, uncomfortable to wear
Long-haired accuracy
<60%
Battery
3–5 days
Comfort
Poor
Technical limitations cap real-world usefulness
C
IMU + temperature + GPS fusion
Recommended path
BCG-based heart rate reference unaffected by fur
Multi-sensor fusion covers health + tracking
14–21 day battery
IP67, every breed
BCG heart rate is reference-grade, not medical-grade
Requires Edge Impulse behavior model training
Multi-sensor integration is more complex
Every breed
Yes
Battery
14–21 days
Differentiation
High
Best balance: health + tracking + battery
Final decision: multi-sensor fusion + zero in-house algorithms
We chose Path C as the core architecture. A Bosch BMI270 6-axis IMU handles BCG heart-rate reference and behavior classification; an ADI MAX30208 high-precision temperature sensor provides ±0.1°C continuous monitoring; a Quectel BG95-M3 module delivers LTE-M/NB-IoT cellular positioning and dual-mode GNSS tracking. On the algorithm side we took a 'zero in-house' approach — step counting via ST FSM built-ins, behavior classification trained on Edge Impulse, GPS using Quectel's reference code — minimizing R&D risk and time-to-MVP.
§ 03 / System Architecture
From sensor to user's fingertip — the data pipeline
The system runs as a modular five-stage architecture: capture, edge inference, data uplink, cloud processing, user delivery. Each stage upgrades and degrades independently — keeping the system maintainable and extensible.
01
Sensor capture
The BMI270 6-axis IMU samples acceleration and angular velocity at 50Hz, the MAX30208 reads temperature every 30s, and the BG95-M3 gets GPS coordinates on demand.
BMI270MAX30208BG95-M3
02
Edge inference
The nRF52840 runs an Edge Impulse behavior classification model locally (walk / run / sleep / still / anomaly). ST FSM computes step counts — cutting cloud upload volume.
nRF52840Edge ImpulseST FSM
03
Data uplink
BLE 5.0 syncs to the phone app in real time. LTE-M/NB-IoT cellular goes straight to the cloud when the phone isn't around. The dual-channel approach guarantees no data loss.
BLE 5.0LTE-MNB-IoT
04
Cloud processing
AWS Amplify receives the data stream. Lambda functions run anomaly detection rules. Amazon Location handles geofence events and triggers real-time push notifications.
AWS AmplifyLambdaGeofence
05
User delivery
The cross-platform uniapp shows health dashboards, GPS maps, behavior timelines, and trend reports. APNs/FCM push anomaly alerts. Multi-member family sharing is supported.
uniappAPNsFCM
Diagram: four-layer architecture — hardware → cloud → app → AI analytics
§ 04 / R&D Process
18-week MVP roadmap
From kickoff to MVP Demo Day, 18 weeks to cover hardware prototyping, firmware, app development, cloud deployment, and full-device validation. Every stage ships with clearly defined deliverables and acceptance criteria.
Total cycle
18 weeks
Core team
Hardware + Software + Product
Key milestones
7
Acceptance
End-to-end MVP demo
Week 1
Kickoff
Complete requirements analysis, competitive research, and tech selection. Lock in the IMU + temperature + GPS fusion path.
Requirements docCompetitive analysisTech selection plan
Week 2–4
Hardware prototype design
Schematic design and PCB layout complete. Lock in the nRF52840 + BMI270 + MAX30208 + BG95-M3 core chipset.
SchematicPCB Layout v1BOM
Week 5–6
PCB fab + firmware development
PCB v1 boards come back. Start embedded firmware in parallel — implement sensor drivers and the BLE protocol stack.
PCB v1 samplesFirmware alphaBLE protocol doc
Week 7–9
Software dev and integration
uniapp mobile development. AWS Amplify backend stand-up. Hardware-software-cloud three-way integration test.
App v1.0 RCBackend APIIntegration test report
Week 10–12
EVT and mold tooling
Engineering Validation Test (EVT) — verify sensor accuracy, battery life, and waterproofing. Start injection mold tooling in parallel.
EVT test reportMold designEdge Impulse model v1
Week 13–15
DVT and full integration
Design Validation Test (DVT). Full device assembly. IP67 waterproof testing. Battery life confirmed at the 14–21 day target.
DVT prototypeWaterproof test reportBattery life data
Week 16–18
MVP Demo Day
Full-featured MVP demonstration end-to-end: real-time health monitoring, GPS tracking, behavior classification, and anomaly alerts.
MVP deviceDemo presentationTechnical doc package
§ 05 / Results
Core technical metrics
Numbers below come from EVT/DVT-stage real measurements, compared against industry baselines — showing the technical advantage in key dimensions.
Metric
Target
Industry baseline
Improvement
Temperature accuracy
Target±0.1°C
BaselineIndustry avg ±0.5°C
Improvement5× better
Behavior classification accuracy
Target92%+
BaselineEdge Impulse baseline 78%
Improvement+14pp
GPS time-to-fix
Target<30s
BaselineCold start 60–90s
Improvement-50%
Battery life
Target14–21 days
BaselinePPG solutions 3–5 days
Improvement4× better
Waterproof rating
TargetIP67
BaselineCompetitors mostly IPX6
ImprovementFully submersible
Every-breed compatibility
Target100%
BaselinePPG <60%
Improvement+40pp
BCG-based heart rate cleanly resolved measurement on long-haired breeds. Not relying on optical sensors means Golden Retrievers, Samoyeds, and similar breeds can finally get continuous health monitoring.
Key acceptance milestones
Schematic review passedWeek 2
PCB v1 sample validationWeek 5
Software v1.0 RCWeek 9
EVT passedWeek 12
DVT device validationWeek 15
MVP Demo DayWeek 18
§ 06 / Use Cases
From daily companionship to professional care
Multi-sensor fusion isn't built for a single use case — it covers every corner of pet life: daily health monitoring, outdoor tracking, family sharing, and professional boarding supervision.
Daily health monitoring
24-hour continuous tracking of temperature, activity, and sleep quality. AI behavior classification flags anomaly patterns and warns owners before symptoms appear — so vet visits happen on time.
TemperatureActivity trackingSleep analysisAnomaly alerts
Real-time GPS tracking
LTE-M cellular + GNSS dual-mode positioning — works even without a phone nearby. Set a geofence and get instant alerts when your pet crosses it. Full track replay records every walk.
Real-time locationGeofenceTrack replayBoundary alerts
Family sharing
Multi-member family accounts share the same pet's health data and location. Whoever's walking the dog or traveling, the whole family stays in the loop.
Family sharingMulti-device syncHealth reportsRemote care
More use cases
Exercise management
Daily activity goals tailored to breed, age, and weight. Track step count and active time to keep your pet at a healthy weight.
Pet boarding monitoring
While your pet is boarded, check activity status and location remotely — making sure the boarding environment is safe and reliable.
Senior pet care
Continuous monitoring for senior pets — temperature swings and reduced activity get caught early, surfacing chronic-condition signals.
Multi-pet households
One app, multiple collars. View all your pets' health and location data in a unified place.
§ 07 / Our Capability
Full-stack IoT delivery
Our team covers the full stack from hardware design to cloud deployment — taking an IoT product from concept to production independently.
Full-stack hardware design
From schematic design and PCB layout to embedded firmware. Full IoT hardware productization. Hands-on with Nordic, Quectel, Bosch, and other mainstream chip platforms.
Edge AI deployment
TinyML model training and deployment on Edge Impulse — bringing real-time behavior classification and anomaly detection to Cortex-M4-class MCUs.
Cloud architecture design
Full-stack AWS Amplify / Lambda / IoT Core design and implementation. IoT data pipelines and real-time alerting systems built to support millions of devices.
Cross-platform app development
uniapp cross-platform mobile development — one codebase covers iOS and Android. Strong on health data visualization, map integration, and real-time communication.
Related solutions
More Wavesteam solutions
AI, capital-markets docs, OCR, vision, IoT and membership operations — composable for your industry.
AI Health Operating System
An operating system for AI health brands — gut testing, report interpretation, personalized intervention, nutritionist workflows, and a repeat-purchase loop in one platform.
- AI
- Health Care
- Mini Program
Coupon & Membership Platform
Issue, redeem, and settle coupons across chain stores and retailers — membership cards, real-time analytics, and finance-ready reporting in one platform.
- Mini Program
- Membership
- Marketing
Capital Markets Doc AI
Bilingual term checks, version diffing, and workflow automation for prospectuses, annual reports, and offering circulars — review cycles cut from weeks to days.
- AI
- Capital Markets
- Automation
Let's Talk
If you have a concrete workflow AI hasn't solved yet, let's figure out the right approach together.
We unpack the workflow with you, judge whether AI is worth using and which approach makes the most sense, then come back within 5 business days with a practical initial plan and estimate.
Business email
contact@boilingwater.cnOffice
10F, South Tower, Kingkey Yujing Times, Longgang District, Shenzhen