To meet these objectives, I contributed to developing My Friend, an AI-powered voice assistant, using OpenAI, ReactJS, NodeJS, ExpressJS, AWS, and Postgres. We focused on integrating AWS services and implementing real-time communication to enhance the system’s performance and user experience.

AWS Service Integration

• Utilised AWS Transcribe for speech-to-text conversion, allowing the system to accurately understand children’s spoken words.
• Integrated AWS Polly to generate natural-sounding voice responses, making interactions more engaging and accessible for young users.
• Used DynamoDB and Postgres for managing both session data and longer-term contextual information. Children’s responses were saved in a vector database, maintaining conversation continuity and enabling personalised interactions over time.

Real-Time Communication

Implemented WebSockets for real-time interactions between children and the voice assistant, ensuring seamless communication without delays.

Built a system that supports real-time guidance, helping children through problems in an interactive and dynamic way.

Created an engaging 3D character, similar to Talking Tom, using Three.js and React Three Fiber. This interactive character speaks and listens like a cartoon, adding a visual, fun element to the platform. The character reacts to the child’s voice inputs, animating and speaking back in real time, making the entire experience more immersive and enjoyable for children.

Architectural Optimization

Optimised the platform’s architecture for scalability and efficiency, ensuring smooth performance as the number of users grows.

Leveraged AWS S3 for secure storage of audio files and conversation data, minimising latency and improving the overall system reliability.