The researcher training AI to detect depression earlier

18 January 2026

Auckland Bioengineering Institute, Health and medicine, Alumni, Staff news

Dr Kunal Gupta knows firsthand what sustained pressure can do to young people. Growing up in India, he was one of an estimated ten million teenagers competing for places in engineering degrees. It started when he was just 14.

"There were long work hours, not enough sleep, educational pressure, and expectations around success. And it was competitive, so you felt you couldn't talk to your friends about what you were going through."

Stress and anxiety were common, but rarely discussed.

"People pushed through, even when they were struggling. Seeking help often felt like something you did only when things became unmanageable, not early on."

Gupta succeeded academically; an engineering degree and then industry roles focused on how people interact with technology, including a startup company developing smart shoes to help visually impaired people navigate, and another involving older people, fall-risk, and exercises to improve balance.

If a system doesn't fit how people actually think, feel, and behave, it won't help, no matter how advanced it is.

Dr Kunal Gupta Auckland Bioengineering Institute, Waipapa Taumata Rau

Projects were co-designed with users and tested in real-world settings. Seeing how users responded to the technology reinforced Gupta's interest in building tools that had tangible, human impact, he says.

"It became very clear: if a system doesn't fit how people actually think, feel, and behave, it won't help, no matter how advanced it is.

"Talking directly with users, particularly those with health or accessibility challenges, showed me how differently people experience stress and wellbeing, and how easily one-size-fits-all tools miss the mark."

And an idea started to form: maybe smart technology could be trained to provide personalised help for young men - men like himself - struggling with their mental health, reluctant to seek help, and poorly served by the services available.

"In my mid-twenties, I experienced depression. I knew something wasn't right, but I couldn't clearly explain what I was feeling. When I eventually sought professional help, the conversation felt rushed and structured around a checklist. I was given answers, but I didn't feel understood.

"That experience stayed with me because it revealed a bigger issue: many people struggle long before they have the words, confidence, or clarity to describe what's happening to them."

Increasingly, devices like smart watches and phones were able to read physiological signals - heart rate, brain waves, blood pressure, breathing, even sweat gland activity, which is often triggered as a "fight or flight" response.

Gupta wondered: What if this technology and the information it supplies could help identify early signs of stress or emotional difficulty, without relying solely on people being able to explain how they feel?

In 2019, Gupta came to New Zealand to do his PhD with Professor Mark Billinghurst and the Empathic Computing Laboratory research group at Waipapa Taumata Rau - The University of Auckland's Bioengineering Institute. His work explored the use of measurable physiological biosignals such as heart rate variability, brain waves, and pupil diameter to design virtual systems that can interpret emotion and respond in real time to someone's emotional and cognitive state.

Take the Tōku Hoa ('My Friend") app development project, which Gupta joined in 2024. The research involves designing and piloting a personalised AI-powered virtual reality buddy who can read someone's biological distress signs, link it with what's going on in their life, and reach out.

"Stress levels are high. Wanna check in?" Tōku Hoa tells a depressed-looking young man in the project concept video, before prompting him to do some stretches, set goals, log mood and stress, and finally get out of the house.

"You have been feeling angry most of the day," he tells another young man, encouraging him to do more steps ("200 more to go" for his daily goal), drink more water and increase his push-ups by ten.

Tōku Hoa catches up again at the end of the day.

"You started the day feeling low, skipped social plans and had a tough morning. But you set a goal for movement and followed through with a workout. Your stress levels stabilised in the afternoon. Small steps, big wins!"

Rather than acting as a diagnostic tool, Tōku Hoa is being developed as an early-support system, Gupta says, drawing on data from smart watches and user input to help people notice patterns over time.

"In simple terms I design technology that understands humans and adapts itself to provide more personalised and meaningful interactions."

The goal isn't to replace human care, but to support people earlier, helping them understand what's happening before things reach crisis point.

The prototype is ready, Gupta says. The next step is to test the system with young men experiencing mental health challenges; Gupta has been granted a Lottery Health Postdoctoral Fellowship to take the prototype into clinical evaluation and community-based settings.

"I've also been awarded a [$360,000] Marsden Fast-Start grant to explore how emotional memory recall in virtual reality, inspired by Māori stories, could support the earlier detection of depression."

Both the Tōku Hoa and memory recall projects are recruiting for masters students.

Dr Kunal Gupta spoke to Milly Smyth for bFM's Ready Steady Learn research segment. Listen to the interview.

Media contact

Nikki Mandow | Research communications
M: 021 174 3142
E: [email protected]

AI at ABI