COMMUNITY ENGAGEMENT

40

Finalist teams

20

Structured consultations

‍

12

Workspaces for organisations

Accelerating Generative AI for Mental Health Research with MEXA

This case study explores Gooey.AI’s contribution to the MEXA Accelerator, a four month program advancing generative AI research for mental health in partnership with Neuromatch and funded by the Wellcome Trust.

It is meant for organisations in the field of mental health research and universities who are considering adopting AI.

COMMUNITY ENGAGEMENT

40

Finalist teams

20

Structured consultations

‍

12

Workspaces for organisations

Wellcome Trust is a global charitable foundation established in 1936, working towards a healthier future for everyone. As a global foundation, the Wellcome Trust has a wide-reaching influence across health research institutions, universities, and policy ecosystems worldwide.

NeuroMatch is a nonprofit organisation dedicated to training the next generation of computational research leaders through inclusive, collaborative, and globally distributed programmes. Their international reach enables researchers across institutions and geographies to participate in shared learning and cross-disciplinary collaboration at scale.

MEXA is a community at the intersection of mental health and artificial intelligence co-created by Wellcome Trust and Neuromatch. MEXA creates real solutions for mental health by leveraging generative AI in an evidence-based and rigorous way.

Theory of Change

The need for trustworthy AI in mental health practice

The gap between AI potential and mental health practice

Generative AI holds significant promise for mental health research, and practitioner support. However, most mental health practitioners and organisations lack the technical infrastructure, capacity, and frameworks required to quickly and effectively use generative AI. As a result, AI adoption by mental health professionals remains limited and does not reach patients as end beneficiaries.

Ethical and privacy constraints

Mental health organisations and institutions require AI partners they can trust, particularly around privacy and concerns that patient data will not be used for model training. Many existing AI tools have inherent biases and ethical concerns, making it difficult to approach AI with ease. Organisations also need awareness of how their peers are using AI responsibly. There is an urgent need for trusted, privacy-first AI platforms in the space of mental health.

A shared platform to support the mental health AI ecosystem

To fund practitioners to use AI in mental health, Wellcome Trust in collaboration with NeuroMatch and MEXA made a $100M+ investment in the MEXA Generative AI for Mental Health Research Accelerator. Without shared infrastructure, however, fundees would have limited cross-project learning and impact. Gooey.AI provides a common, safe, privacy-first platform that allows funded teams to rapidly build, evaluate, and deploy AI tools quickly.

Background and Needs

Lack of AI expertise

Wellcome Trust funded this Accelerator to support mental health researchers and practitioners with domain expertise but limited experience in generative AI. While the programme provided mentorship, and funding to mental health research and AI model development, teams still needed a trusted AI partner. The 40 finalist teams lacked AI expertise and needed collaborative support to help bring their research ideas to fruition and turn them into functional AI models.

Need for upskilling

The core aim of the accelerator is to enable teams to produce high-quality research proposals at the intersection of AI and mental health, positioning them for Wellcome Trust’s funding call of up to £3 million per award. MEXA-led workshops covered experimental design, ethics, and lived experience integration. However, effective upskilling also required hands-on access to AI platforms to translate this learning into practice. This is where Gooey played a critical role.

Projects for anxiety, depression and psychotic disorders

Accelerator projects focused on complex mental health conditions like anxiety, depression and psychotic disorders that demand care and an ethical approach between AI systems, clinicians, and people with lived experience. For this, teams required a trustworthy tech partner who could safely work with sensitive data, evaluate models, and iterate across diverse languages.
‍
An AI partner such as Gooey provides the shared, privacy-first infrastructure needed to support this work, enabling teams to focus on research rather than these limiting constraints.

Interventions & Process

Protecting private data

Gooey created a shared knowledge hub to surface common problems, and ethical concerns across the cohort. We also conducted 20 structured consultations to help teams translate mental health research goals into AI workflows.

Participants were able to prototype their AI workflows using zero data retention, and VPC or on-prem deployment options. We ensured that patient and research data was never used for model training.

Hosting & evaluating fine-tuned models

We created workspaces for 12 organizations for rapid prototyping and evaluation. Teams could build golden datasets, evaluate competing state-of-the-art (SOTA) models on Gooey, and iteratively improve their working models.Once models were ready, they could deploy securely via Gooey endpoints, cloud, VPC, or on-prem infrastructure.

Understanding low-resource languages

As a platform designed to understand any language, we could support the effort of MEXA teams working in multilingual and low-resource contexts.

Teams could evaluate which SOTA models best understood their users’ languages, using bespoke evaluations based on user audio and transcriptions. Both open and private models were researched, hosted, and evaluated by Gooey to ensure appropriate model selection by the MEXA participants.

Deploying to comms. channels

To support real-world testing and usability, Gooey helped MEXA teams deploy prototypes across accessible communication channels. Teams connected AI agents to WhatsApp Biz Provider, Voice + SMS, Skinable web widgets. Integrated analytics and dashboards were beneficial to teams to gather insights from end beneficiaries.

Read more about our process here.

Outcomes

Shared infrastructure for learning

The MEXA Accelerator is a four-month programme designed to support interdisciplinary research teams with resources, mentorship, and strategic guidance to develop proof-of-concept projects in generative AI for mental health.

Gooey strengthened this community by creating shared knowledge hubs, hosting consultations, and providing secure workspaces. Our 24/7 support enabled more than a dozen teams to submit high-quality research applications to the accelerator.

A supportive cohort advancing AI research for mental health

We played a central role in fostering a collaborative and supportive cohort within the MEXA Accelerator. Teams could share insights, iterate rapidly on AI workflows, and learn from one another’s experiences while using our platform.

Our privacy-first infrastructure allowed participants to work safely with sensitive mental health data. Together, our interventions helped to build a cohesive and thriving community of researchers advancing generative AI for mental health, with accelerator winners to be announced in February 2026.