I'm Building an AI Neonatal Monitor for the Hackster.io Arduino UNO Q Contest
Every year, approximately 2.5 million newborns die in the first 28 days of life. The overwhelming majority of these deaths happen in low-resource settings — district hospitals and rural clinics across sub-Saharan Africa, South Asia, and Southeast Asia — where a single nurse may be watching over six to ten newborns simultaneously with no electronic assistance.
A clinical-grade neonatal monitor that could prevent many of these deaths costs between $10,000 and $15,000. That price puts it completely out of reach for the hospitals that need it most.
I'm building NeoGuard to change that.
The Contest
I've submitted a hardware application to the Hackster.io "Invent the Future with Arduino UNO Q and App Lab" contest, organised by Arduino, Qualcomm, and Edge Impulse. The contest offers a $20,000 prize pool and provides free Arduino UNO Q boards to selected applicants.
NeoGuard is targeting the Social Impact prize.
What is NeoGuard?
NeoGuard is an AI-powered smart neonatal bassinet designed for deployment in low-resource clinics. The baby lies inside the device. Sensors embedded in the walls and floor monitor vital signs continuously and automatically from the moment the baby is placed inside.
What it monitors:
- Heart rate and blood oxygen (MAX30102 optical sensor)
- Electrical heart rate cross-validation (ADS1115 + ECG pads via isolation amplifier)
- Breathing and movement (MPU-6050 IMU under the foam mattress)
- Non-contact skin temperature (MLX90614 infrared)
- Ward air quality — temperature, humidity (BME280)
- Visual breathing confirmation (MIPI-CSI overhead camera)
What makes it different from a simple monitor:
- AI eliminates false alarms through multi-sensor cross-validation — both HR channels must agree before an alarm triggers
- Edge Impulse anomaly detection on-device confirms apnea only when multiple sensors agree
- Shift handover reports generate automatically at shift end
- Built-in Wi-Fi hotspot — nurses connect a smartphone or tablet directly, no router or internet required
- Voice alerts in the local language via HuggingFace TTS, running fully on-device
- Total hardware cost: approximately $150
The Hardware: Arduino UNO Q
The Arduino UNO Q is what makes this project possible in its current form. It's a dual-core board combining:
- STM32U585 (Cortex-M33) — the real-time microcontroller side, running bare-metal C for sensor acquisition with alarm latency under 5ms
- Qualcomm QRB2210 (Cortex-A53) — running Debian Linux for AI inference, camera processing, and the App Lab clinical dashboard
These two processors communicate via Bridge RPC over a UART serial channel. The STM32 reads all sensors and drives the alarm hardware independently. The Linux side handles intelligence, the dashboard, and voice alerts. If the Linux side fails, the STM32 continues operating and alarming on its own.
This dual-core architecture is exactly what a medical device needs: hard real-time guarantees on the safety-critical path, and a full Linux environment for the smart features.
The Physical Design
The bassinet is a 3D-printed oval shell (60×30×20cm) in white PLA/PETG, elevated on four short legs for airflow. A transparent acrylic dome encloses the baby, maintaining temperature and reducing infection risk.
A solar panel sits on top of the dome for off-grid deployment. A LiPo battery with TP4056 charging and a 3.7V→5V boost
Why I'm Sharing This Now
I believe in building in public. Every two weeks I'll publish an update on both this blog and the Hackster project page, showing real sensor readings, firmware commits, and physical build progress.
Hackster project page: hackster.io/contests/invent-the-future-with-arduino-uno-q-and-app-lab/hardware_applications/18952
GitHub: github.com/FrankApala