Work / NELA Cough: AI tele-monitoring for chronic-disease patients, from requirements to series production.Edge AI for remote patient monitoring
Context
I spent almost three years at NELA (Carepath Technologies GmbH), building NELA Cough, a system for tele-monitoring and self-management of patients with chronic diseases. I joined as an AI Software Engineer and left as Senior IoT Engineer, working on the product from the first model to production hardware used by patients.
The work involved IEC 60601 requirements, traceability and clinical stakeholders in a regulated medical device environment.
The problem
The product has to detect and characterise a cough from a device in a patient's home. The audio cannot be sent continuously to a server, so the model has to run predictably within the MCU's limited compute budget. Patients cannot be expected to debug it.
Clinical requirements arrived as prose, and hardware suppliers' timelines could change. The device also had to meet certification requirements.
What I did
- Led development of AI models for medical applications, and their real-time deployment on edge devices and MCUs.
- Owned the IoT system architecture end to end: requirements, hardware and software design, and firmware, from prototype to series production.
- Built the cloud back end on Azure IoT, and production firmware for the devices.
- Worked directly with customers, partners and external medical hardware suppliers on integration and deployment issues, and wrote the demos and technical documentation to support that work.
Stack & notes
Python · TensorFlow · Azure IoT · embedded systems · RTOS · IEC 60601.