Melbourne-born health tech startup Heidi has had a surprisingly rocky start for a company that has saved doctors around 43 million hours of administrative work.
Six years ago, co-founders Yu Liu and Dr. Thomas Kelly built a medical chatbot to train medical students on how to consult with patients. When that didn’t gain traction, it pivoted to a general practitioner (GP) platform that offers telemedicine, medical consultation bookings, and prescriptions.
“We learned a lot from that experience,” Heidi chief technology officer and co-founder Yu Liu told Computer Weekly in a recent interview, adding that the company initially struggled to sell its software to clinics that were already using more established applications.
Although customers who adopted Heidi’s GP platform saw productivity increases of up to 20%, it wasn’t enough for Liu and his team. “My belief is that if you want to change something that’s currently working, you need to improve it by more than 100 percent,” he says.
This realization led to a second pivot towards automating medical consultation transcription and clinical documentation, which is at the core of the business today. Currently valued at $465 million and backed by $100 million in total funding, Heidi’s cloud-based platform is used in over 110 countries and processes over 2 million consultations each week.
In February 2026, the company acquired AutoMedica, a UK-based clinical AI pioneer, expanding its footprint in Europe. We also launched a series of new tools: Heidi Comms for automated patient outreach and Heidi Evidence, an ad-free clinical research tool built in partnership with medical databases and journals such as BMJ, to provide physicians with trusted, verifiable research in the clinical setting.
Heidi’s most ambitious move is her foray into hardware. The company recently launched Heidi Remote, a 21g wearable AI microphone specifically designed for the high-decibel, chaotic environment of hospitals.
“I think within the next three to five years, everyone will have a remote AI wearable,” Liu said. “Right now, when you’re walking around, a significant portion of the clinical session is lost because your microphone is turned into a conference mic or you have to hold your phone in your hand, which is very inconvenient.Also, doctors don’t always have good Wi-Fi.”
The device captures and automatically syncs audio offline and boasts a 14-hour battery life. But for Liu, this hardware is a stepping stone to a broader vision of eliminating computers from exam rooms completely.
“About 30 percent of a doctor’s time is spent inputting what they’ve gathered in their head into a computer,” Liu explains. “We want to eliminate the use of a computer and combine it with a voice interface that turns prompts into tasks. We can say, ‘I’m done with this patient. Check out this BMJ article for dosages. If that’s OK, open your prescription software and put in the medication.'” After a minute, the doctor can see and confirm everything. ”
To power this level of complex automation, Heidi has changed the way it builds artificial intelligence (AI) capabilities. The startup initially relied on off-the-shelf basic models, but high latency, rising costs, and lack of clinical optimization led them to build their own models.
Currently, more than 80% of Heidi’s workloads run on its own model. The startup uses a proprietary blind A/B testing system to present two AI-generated outputs side-by-side to doctors and objectively measure which model performs better in a real-world clinical setting.
However, Heidi has not completely abandoned the frontier model provider. We maintain a close partnership with Anthropic, using our Claude models for the complex inference tasks Heidi Evidence performs, prototyping new features, and generating synthetic data to fine-tune our models.
As the company expands, with cumulative consultation volume up 20% month over month in markets such as France, Germany, and Singapore, Liu is focusing on an infrastructure strategy that moves AI away from the public cloud.
Recognizing that data privacy remains a major barrier to large-scale hospital adoption, Heidi is developing an on-premises strategy where the entire AI software suite is packaged onto a physical server box and shipped directly to healthcare facilities.
“Nobody wants to send their data somewhere else,” Liu said. “Hospitals can start using Heidi and no data will leave the hospital’s site. We believe we have at least a 6-12 month advantage over our competitors in delivering this.”
