Inspiration
Growing up, I witnessed friends and family in remote regions struggle to access timely medical care. In places where a clinic is hours away and reliable internet is spotty at best, even changes in blood-oxygen levels or heartbeat can go unnoticed until it’s too late. We wanted to create something that brings hospital-grade vigilance to rural communities: something lightweight and affordable that works off-grid. That spark became DeepLink, an AI-driven wearable that connects hearts to doctors, loved ones, and emergency services, no matter where you are.
What it does
- Always-On Monitoring
Continuously tracks vital signs (O₂ saturation, heart–rate trends) using low-power sensors. - Edge-AI Triage
An ML model on the belt classifies readings into Healthy or detects conditions like Hypoxemia, Tachycardia, Bradycardia, HeartSpike, O₂ Anomaly, or Sleep Apnea every few seconds even without internet. - Seamless Rural Connectivity
Uses LoRa protocol to hop data across village gateways, and Starlink backhaul to the cloud, so no one is left offline. - Multi-Tier Alerts
- Stable → periodic “Stable” ping to the doctor’s dashboard
- Caution → instant SMS to family or carers
- Critical → automated voice call to *911 *
- Stable → periodic “Stable” ping to the doctor’s dashboard
- Doctor Oversight & Historical Trends
All data streams into a lightweight dashboard where physicians can review real-time charts, set custom thresholds, and analyze week-long patterns for early intervention.
How we built it
Hardware & Dev Boards
- Micro:bit as an ultra-low-power proof-of-concept and demo platform
- ESP32 for Wi-Fi prototyping and real-time data uplink
- Arduino for connecting smart home devices to the cloud
- Micro:bit as an ultra-low-power proof-of-concept and demo platform
Connectivity Layer
UART → LoRa: Belt MCU sends data via UART to the LoRa radio for long-range transmission.
LoRa Relay: Gateways receive data via LoRa and forward it onward across the network.
UART → WiFi: Gateways take the LoRa-received data and pass it via UART to WiFi for internet/Starlink backhaul.
AI & Software
- Languages & Frameworks: C#, JavaScript, PHP, C, Python, Wolfram, MySQL
- ML Stack: TensorFlow, NumPy, scikit-learn
- Model Architecture: LSTM-based time-series predictor
Backend & Dashboard
- API & Server: Pure PHP backend handling authentication, ingesting LoRa data, and triggering alert workflows.
- API & Server: Pure PHP backend handling authentication, ingesting LoRa data, and triggering alert workflows.
Challenges we ran into
- Hardware Debugging
- Communication Protocols: Orchestrating UART↔LoRa↔UART↔WiFi handoffs
- Interoperability: Harmonizing diverse dev-boards (ESP32, Arduino, Micro:bit) and their firmware stacks for smooth end-to-end data flow.
Accomplishments that we're proud of
- Reliable Communication: Smooth data flow between ESP32, Arduino, Micro:bit, LoRa protocol, and PHP server during our demo.
- Fully Functional Emergency Alert Assistant: Automated 911 calls are triggered accurately on critical events.
- Interactive User Interface: A responsive PHP dashboard for doctors and families, optimized for low-bandwidth environments.
- High-Accuracy AI Model: LSTM-based predictor achieving over 90% accuracy for disease risk classification.
What we learned
- How to Use Conductors and Workflows for orchestrating complex, multi-step data processing in cloud environments.
- Connecting Hardware to Software end-to-end, from sensor signal acquisition to real-time web dashboards.
What's next for DeepLink
- Clinical Validation: Initiate pilot studies and clinical trials with healthcare institutions to measure impact and refine AI models.
- Feature Expansion: Integrate sensors and enhance predictive analytics for broader health monitoring.
- Scale to More Communities: Deploy additional gateways across rural regions and partner with NGOs for greater reach.
- Manufacturing & Cost Optimization: Streamline production processes and secure component supply chains for mass manufacturing.
- Regulatory Approvals: Pursue FDA, CE, and other certifications to enable global market entry and healthcare integration.
Built With
- arduino
- c
- c#
- esp
- javascript
- lora
- lstm
- micro-bit
- mysql
- numpy
- php
- python
- scikit-learn
- tensorflow
- uart
- wifi
- wolfram-technologies
Log in or sign up for Devpost to join the conversation.