AxonBuild - AI & Full Stack Solutions

Introduction

We developed a cross platform mobile application that delivers AI driven conversational therapy for common mental health challenges such as sleep disorders, anxiety, and eating habits. The app was built in React Native with Expo and connected to Firebase for authentication and data storage. A FastAPI backend orchestrates retrieval augmented guidance, clinical case analysis, and structured script delivery. By combining evidence based coping strategies with personalized AI conversation, the platform provides both immediate relief and longer term support.

Challenge

Access to mental health resources is often limited, and many individuals lack timely support for everyday struggles such as stress, anxiety, or sleep disruption. Therapy can be expensive, geographically inaccessible, or intimidating for first time seekers. Existing wellness apps typically offer static exercises without context, memory, or personalization. We needed to create a scalable and secure therapeutic assistant that feels human like, adapts to each user's situation, and works seamlessly across devices.

Solution

We designed a comprehensive therapeutic AI platform with multiple integrated capabilities.

Voice Support

Beyond text chat, the app accepts voice notes. Audio is transcribed with Whisper, processed through the same therapeutic reasoning pipeline, and responded to with text and optional audio output.

Clinical Case Analysis

A background analyzer reviews each session to confirm which solutions have been delivered. This intelligence ensures comprehensive coverage of coping strategies across conversations.

Disorder Identification

Through natural dialogue, the chatbot identifies user concerns and symptoms. These are embedded and matched against a library of curated clinical cases in Qdrant, grounding the assistant's reasoning in evidence based examples.

Cross Platform Experience

Built with React Native and Expo for iOS and Android, with Firebase for user management. FastAPI delivers backend APIs, Qdrant handles vector retrieval, and OpenAI models power embeddings, chat completion, transcription, and text to speech.

Retrieval Augmented Guidance

The assistant supplements conversations with guidance notes retrieved from a vector database. This ensures coping strategies are not just generic LLM outputs but context informed suggestions tied to real cases.

Session Memory and Personalization

Each user's session is keyed by profile and email. The system tracks conversation history, identified disorders, offered scripts, delivered solutions, and completed cases. This continuity allows the assistant to avoid repetition and build a sense of therapeutic awareness.

Script Offering and Guided Practices

The chatbot tracks conversation flow and session memory. When a disorder is identified and the timing is right, it delivers a relevant soothing script such as breathing exercises or relaxation steps. These scripts are drawn from a structured library, ensuring users receive the right practice at the right moment.

Technical Approach

Mobile Development:React Native with Expo for cross platform iOS and Android deployment

Authentication:Firebase for secure user management and session data storage

Therapeutic AI:RAG powered guidance using clinical case embeddings in Qdrant

Voice Processing:Whisper transcription with text to speech response capabilities

Session Management:Personalized conversation history and disorder tracking

Clinical Analysis:Background case analysis ensuring comprehensive therapeutic coverage

Stack:React Native frontend, Python FastAPI backend, Firebase database, Qdrant vector store

What we've accomplished

Successfully delivered a comprehensive cross platform therapeutic AI companion that provides accessible mental health support through evidence based conversational therapy. The solution combines clinical expertise with modern mobile technology to deliver personalized, scalable mental wellness support.

Results & Impact

•

Accessibility:Therapeutic support delivered directly on mobile without cost or geographic barriers

•

Personalization:Conversations adapt to user profiles, symptoms, and history across sessions

•

Engagement:End of session scripts increased user retention and session completion rates

•

Scalability:Firebase, FastAPI, and Qdrant supported efficient deployment and multi user growth

Project Narrative

The project began with simple relaxation scripts packaged in a mobile app. It evolved into an intelligent assistant that identifies user concerns, retrieves clinical guidance, and offers both immediate relief and structured long term strategies. Adding voice input and transcription made the assistant accessible to users who prefer speaking naturally. Script orchestration, session memory, and background clinical case analysis ensured conversations stayed relevant and supportive over time. What started as a static wellness tool matured into a comprehensive therapeutic companion that feels human like, adapts to user needs, and provides trusted support on any device.