PsyAI

The project

PsyAI is a mental health support platform designed to bridge the gap between therapy sessions. Many patients only see their therapist periodically, making it difficult to reflect on their thoughts and emotions in between appointments. Our goal was to provide users with a safe and accessible environment where they can journal, track their mood, and engage in structured CBT-inspired reflection, while enabling therapists to gain relevant insights into patient progress.

At the core of the platform is an AI-powered chatbot that supports structured reflection through CBT-informed conversations. The chatbot uses a Retrieval-Augmented Generation (RAG) pipeline grounded in Cognitive Behavioural Therapy principles to improve response relevance and consistency. To minimise unsafe or inappropriate model behaviour, the system includes multiple safety layers such as input/output filtering, crisis detection, and rule-based guardrails.

The rest of the platform is built as a modular full-stack system consisting of a Next.js frontend, a FastAPI backend, and a Supabase database with Row Level Security for secure data storage. The AI chatbot is integrated as a separate service that communicates with the backend, ensuring a clear separation between application logic and model inference.

By combining guided conversations, journaling, and therapist connectivity with an AI chatbot for guided reflection, PsyAI provides continuous support between therapy sessions while maintaining clear boundaries between AI assistance and clinical care.

Supporting therapy doesn't mean replacing therapists — it means making the time between sessions more meaningful.

The customer

Our client was the founder of PsyAI, who envisioned a platform that could provide meaningful mental health support between therapy sessions. While working closely with psychology researchers, the client aimed to explore how AI could be applied responsibly within mental healthcare. During the project, we transformed this initial concept into a functional prototype that demonstrates the platform’s core features and AI capabilities.

Communication with the client was regular and collaborative, consisting of meetings, feedback sessions, and demos. This helped us refine the product and ensure alignment with the intended vision. The final result provides a foundation for further evaluation and potential testing with psychology researchers.

Turning an idea into a working product is just the first step, the real impact begins when it's put into practice.

The team

Our team worked using Scrum, with roles including Scrum Master, Product Owner, frontend development, backend development, AI development, and infrastructure. While responsibilities were divided, we worked closely together through sprint planning, code reviews, and continuous integration to ensure all components worked together.

During the project, we faced challenges such as integrating AI components, managing authentication and role-based access control, and designing a privacy-conscious architecture. Combining multiple systems required strong coordination between team members.

As a team, we are most proud of delivering a complete end-to-end platform that combines modern web development with responsible AI design.

Working with 6 developers, each with their own ideas and schedules, taught us that alignment is just as important as implementation.

The technologies

Languages:

• TypeScript — frontend development
• Python — backend services and AI pipeline
• MySQL , PostgreSQL — database design and data management

Frameworks & Platforms:

• Next.js (React) — frontend application
• FastAPI — backend API
• Tailwind CSS — user interface styling
• Supabase — authentication, PostgreSQL database, and Row Level Security (RLS)

Artificial Intelligence:

• Retrieval-Augmented Generation (RAG) — grounds AI responses in CBT knowledge
• Local Large Language Models (Gemma via LM Studio/Ollama) — CBT-informed conversational AI
• BERT — crisis and suicide risk detection

Tooling:

• GitHub — version control and collaboration
• ESLint — code quality and consistency

Deployment:

• Containerised architecture with a Next.js frontend, FastAPI backend, Supabase database, and locally hosted LLM.