Service
App development, AI integration
Psyvatar AI
About Psyvatar AI
Psyvatar's CEO Justin Vezeau approached DETL Inc with the ambitious goal of building a mental health AI app where users can talk to their AI therapist. With the recent rise of AI and its controversy behind people talking more to robots and less to human beings, developing a mental health app where users talk to an AI therapist doesn't help reduce this stigma either. And thus the goal was to develop an mental health AI app where users are talking to an AI therapist but won't be able to distinguish between talking to a human vs talking to an AI therapist.
DETL Inc accepted this challenge to create this application. To make this project even more challenging, Justin also emphasized a tight deadline. The app needed to be completed within a 1.5 months time frame in order for BETA users to test the app and eventually launch on the Apple and Google Play store.
The Description
The Psyvatar app required the implementation of various features in order to enhance the authenticity of user - therapist interactions when the user is in a session with the therapist. The following features were of paramount importance:
- The therapist should recall previous sessions
- The information exchanged between the AI and the user needs to be anonymous.
- The therapist should identify patterns in order to provide clarity and insight to the user.
- Users should have access to a diverse roster of therapists, allowing them to browse and select the professional that best suits their needs and preferences.
- To enable real-time, hands-free conversations, the app should incorporate voice detection.
Owner
Justin Vezeau
Project Date
Start: 08/2024 - End: 10/2024
The Challenge
The development of the Psyvatar AI app presented several significant challenges. Firstly, implementing a custom Retrieval-Augmented Generation (RAG) model was crucial to enable the AI therapist to build a deeper, more personalized connection with users. This required advanced natural language processing techniques to handle complex interactions. Secondly, the client set a tight deadline of just 1.5 months to launch the app and make the beta version available to users, demanding efficient project management and rapid development cycles. Lastly, we needed to integrate custom voice detection to allow seamless, hands-free conversations between users and the AI therapist, eliminating the need for users to hold the phone or press a button every time they wanted to speak. This posed technical challenges in ensuring accurate and real-time voice recognition.
The Solution
To address these challenges, we implemented a custom Retrieval-Augmented Generation (RAG) model by transcribing all conversations between users and the AI therapist. These transcriptions are stored in a large, anonymized file accessible to the AI at the start of each session. By removing all personal identifiers, we ensured that the AI could reference past interactions without compromising user anonymity, allowing for a deeper and more personalized connection. To meet the tight 1.5-month deadline, we expedited the development process by completing the UI/UX design within two weeks, maintaining constant communication with the client for iterative feedback and rapid adjustments. For seamless user interaction, we developed a custom voice detection system that monitors the phone's sound levels to detect when the user is speaking. This enabled real-time, hands-free conversations without the need for users to hold the phone or press buttons, enhancing the overall user experience with the AI therapist.
As a result, we successfully developed and launched the Psyvatar AI app within the tight 1.5-month deadline, incorporating all the required features:
-
Comprehensive Authentication Flow: We implemented a seamless onboarding process with complete sign-up and sign-in functionality, enhancing user experience from the very first interaction.
-
Face ID Integration: For added security and convenience, the app supports Face ID authentication, allowing users quick and secure access to their accounts.
-
Notification System: Users receive timely notifications to stay engaged with their therapy sessions and receive important updates, ensuring consistent interaction with the app.
-
Therapist-User Sessions: We enabled seamless real-time sessions between the AI therapist and users, powered by our custom RAG model and ChatGPT integration, providing personalized and meaningful conversations.
-
Session Timer: A timer feature was incorporated to effectively manage and track the duration of therapy sessions, helping users and the system maintain optimal session lengths.
-
Custom RAG Model: The Retrieval-Augmented Generation model allows the AI therapist to reference previous anonymous conversations, building deeper connections without compromising user privacy.
-
ChatGPT Integration Layer: By integrating ChatGPT, we enhanced the conversational capabilities of the AI therapist, delivering more natural and intuitive dialogue experiences.
-
AWS Implementation: Leveraging AWS services, we ensured the app's backend is scalable, reliable, and secure, handling data storage and processing efficiently.
-
Account Management Features: We developed robust account management functionalities, giving users control over their profiles, settings, and preferences within the app.
Overall, the Psyvatar AI app delivers an innovative, secure, and user-friendly platform for AI-driven therapy. It meets all the client's requirements and provides users with an exceptional experience, effectively addressing the challenges and fulfilling the project's goals within the stipulated timeframe.