Introduction
In a night shift on a cardiac ward, Nadia Brem once counted more alarms than meaningful conversations in a single hour. Some alarms mattered, others were noise dressed up as urgency. By morning, nobody trusted the quiet beeping as much as they should have.
Nadia lives in Hamburg and works as a nursing professional coordinating medical AI pilots. She brings the ward into meetings where software teams might otherwise discuss “users” without having seen a handover under time pressure.
Story of the Path into AI
Nadia’s entry point was alarm fatigue. She did not want to become a classical developer; she wanted nursing experience to shape the systems that would later appear on the ward. At first, some doctors and IT staff underestimated her technical curiosity. She, in turn, had to learn to translate clinical observation into data language.
She completed a part-time certificate, learned basic statistics and interviewed colleagues about real workflows. Her first project was an analysis of alarm reports: which warnings were useful, which created stress, and which disappeared because nobody had time to respond. The first dataset looked orderly until Nadia noticed that night-shift notes were shorter and often written later. The model treated that as a pattern, not as exhaustion.
That error changed her way of working. She began to document not only events, but the conditions under which they were recorded.
Current Work
Today Nadia helps embed AI-supported risk notices into nursing practice. In one fall-risk pilot, she insisted that alerts should not appear only in a doctor’s dashboard, but where nurses prepare handovers. The recommendations are checked daily, and false confidence is discussed openly.
The result is modest and important: fewer irrelevant alarms, better acceptance and more realistic expectations. Nadia keeps repeating that clinical AI fails when it ignores shifts, staffing levels and the small workarounds that keep care running. A model can point to risk; it cannot take responsibility for a patient.
Personal Advice
“Anyone building AI for healthcare should walk one shift first,” Nadia says. She does not mean a guided tour. She means the hours when the corridor is too bright, the documentation is late and the person closest to the patient sees a problem before the system does.
Key Facts
Age and place: 28, Hamburg.
Background: shift work, part-time training, interprofessional tension.
Entry into AI: analysis of clinical alarms and their usefulness in daily care.
Focus today: medical AI in nursing practice.
Typical tools: risk models, statistics, clinical process analysis.
Werkstattnotiz
Nadia keeps one alarm log because it looks almost perfect and is therefore suspicious. The times are rounded, the notes are short, the pressure of the night is invisible. She now asks every pilot team: what work produced this data, and what did the person writing it have to leave out?