Wesper: Sleep Lab at Home
Firmware, Cloud Architecture, and Backend Development
I arrived at Wesper in 2020 as the 8th employee and first full-time software hire.
At that time, the company had a functioning demo device & app, but it was not FDA cleared, had no backend, and the product had yet to be sold to a customer.
My Primary Contributions
- Originated cloud architecture, backend API, and database schema/overall cloud dataflow for ingesting & processing sleep data, enabling us to ship to our first customers
- Added firmware support for new sensors, onboard algorithms (i.e. live SpO2 and heart rate monitoring), and reporting capabilities over BLE, resulting in FDA clearance for sleep apnea diagnosis in 2022
- Enabled mass production in China—added firmware support for multiple boards & processors, wrote software for automated board programming, testing, and tracking by ID
Left: patches + v2.8 board; Right: 2 patches are adhered to the skin overnight as shown; this entirely replaces an in-lab sleep test for sleep apnea diagnosis.
Company Growth
When I left in 2022, Wesper had grown to 23 employees and raised a Series A. Now it’s up to 30+ employees and several hundred patients per night.
Tech Specs
Patches:
- NRF52-based custom boards
- Firmware written in bare-metal C with NRF SDK
- Streams data to phone app over BLE; phone relays data to cloud
Cloud:
- All services running on AWS—uses EC2, ECS, Lambda, S3, RDS, etc.
- All backend written in Python (API is in Flask), except the core algorithm suite written in Matlab
Takeaways
- Learned cloud architecture, backend development, and some front-end development independently on the job
- Learned how to ship a consumer hardware product (design-for-manufacture, mass production, agile development & product management, working with designers & C-suite)
- Realized I wanted to change focus to algorithms & machine learning