With more than 1 in 8 adults affected by serious mental illnesses annually and only 7% receiving adequate treatment, psychiatry remains a critical public health challenge. The rising prevalence of mental health conditions, coupled with a growing shortage of clinicians, demands innovative approaches to mental health care.

French startup Callyope is developing solutions to enhance both efficiency and quality of care in psychiatry. Cofounders Xuan-Nga Cao, Rachid Riad and Martin Denais built a specialized speech foundation model that monitors mental health symptoms – like depression, anxiety, insomnia, and psychosis – by integrating patient speech, smartphone sensors and clinical information.

Monitoring brain health with speech foundation models

“Patients download our app and set up their own disease monitoring strategy: symptom self-reporting, voice journaling or passive speech assessment – where our system analyses phone conversations with selected caregivers.” explains Martin Denais. “Just one minute of speech – combined with smartphone sensors and medical data – provides sufficient input to assess most symptoms in depression, bipolar disorders and schizophrenia. This creates an objective monitoring system with minimal burden, and maximal agency for patients”.

At the core of Callyope’s technology is their audio-language foundation model, which processes speech data, sensor metrics related to sleep or activity, and clinical information to generate contextualized clinical insights. This model enables a range of clinical applications – all accessible through Callyope Copilot, the first AI copilot for mental health practitioners.

“Currently, clinicians can use Callyope Copilot to assess symptoms through our CallyoCare MD speech analysis tool, and to summarise complex cases from uploaded medical docs” Martin notes. “We’re advancing our clinical roadmap to enhance our medical device for remote monitoring and relapse detection – with the goal of becoming the essential daily tool for all mental health practitioners.”

AI is bound to revolutionize mental health care

The co-founder acknowledges a key challenge facing AI healthtech companies: ensuring AI models are trained on diverse, real-world datasets to maintain clinical performance outside controlled settings.

Despite these hurdles, he believes AI will become essential in mental health care: “Demand for mental health care is growing dramatically, while we face a worsening global shortage of clinicians. If we hope to improve – or even maintain – care quality, we must develop new approaches to mental health care delivery." Martin says.

Callyope relies on Google Cloud to power its infrastructure and on Gemma open models to develop their deep learning technologies and for a range of internal projects. As participants of the Growth Academy: Google for AI Health program, the founders had the rare opportunity to gain practical advice from the Google team, ranging from Deepmind top AI researchers to Google cloud cybersecurity experts.

Martin says “The startup journey involves numerous technical and personal challenges. Having direct access to expertise from the Google team has been incredibly helpful,” says Martin.

The founders envision Callyope Copilot as the essential cornerstone tool for all mental health clinicians, supporting more informed clinical decisions by seamlessly accessing and interpreting critical data from both electronic health records and patient behavioral patterns.

Today, Callyope is conducting seven clinical studies with over 1,000 patients and 2,000 controls — with plans to publish multiple research papers to drive the field forward.

“We’re just getting started,” says Martin.