Startup Profile
Nixi AI

Introduce yourselves

Tell us in three sentences who you are

Nixi AI is the intelligent layer that has been missing from healthcare: it automates the manual, time-consuming steps of clinical workflows. The physician speaks once, and the documentation, coding and letters are produced and written back into the practice systems, saving time while producing more complete, higher-quality data for prevention, research and health economics. Founded in January 2025.

What challenge does your startup address and how did the idea for it come about?

At a pharmaceutical company, Mehran built a clinical decision support system for diabetologists with a team of twenty. It was recognised by leading clinicians as the best of its kind, and barely used. UX research showed why: time pressure. That became our premise, value only counts if it costs the physician no time.

Three challenges since: building trust in AI among physicians (first paying customers 2026), integrating with the closed legacy systems of German healthcare (PVS, LIS, KIS), and optimising accuracy and latency for German medical terminology.

How does Nixi AI work and how does AI support this project?

Nixi is a modular, agentic platform operated entirely by voice. One agent drafts, one codes, others review and correct before the physician sees anything. The modules in use today are what our customers already pay for; further modules are in active development. The principle stays the same: one voice trigger, and the manual steps run themselves.

How do you ensure data protection?

Privacy by design. All data stays in Germany and the EU, end-to-end encrypted; our providers are bound under §203 StGB, so only the user sees decrypted data. Our own anonymisation engine redacts protected health information with 96% accuracy, so what reaches the model is anonymized.

Which medical specialties is your product compatible with?

Around 68 specialties are supported — among them rheumatology, neurology, cardiology, orthopaedics, gynaecology, paediatrics, psychiatry, anaesthesiology, surgery and general practice. The engine is specialty-agnostic; only the templates differ. It also covers roughly 55 languages, European and non-European: German, English, French, Spanish, Portuguese, Arabic, Hindi, Turkish, Persian and others.

Let’s take a look into the future: where could the project be in five years?

From type-free today to hands-free tomorrow: an agentic voice coworker that captures, codes, writes and acts across every clinical system, while the physician’s hands stay with the patient. And because the data is finally captured completely and structured, it becomes usable for prevention and research, which is why we started.

What is your biggest point of discussion within the team?

Speed versus certainty. The market rewards moving fast; our credibility rests on output a physician can sign unchecked, which you only lose once. Certainty usually wins.

How do you respond to your biggest critics?

The question we are asked most often is whether patients are comfortable knowing that AI is processing their health data. That is precisely what we examined in our validation study, and patients were not merely accepting: they wanted their physician to use it again at the next visit. Our interpretation is that they notice the difference in the consultation itself, because the doctor is more present.