Maraca

Project Overview & Architecture

Maraca is an innovative platform designed to help sales professionals improve their cold calling skills through practice with AI-powered voice agents. Users can create custom personas by defining parameters such as the prospect's designation, availability, temperament (friendly or rude), and other behavioral characteristics. This allows sales teams to practice handling various scenarios they might encounter in real-world sales calls. I developed the entire platform using React.js and TypeScript for the frontend, creating an intuitive and responsive user interface. The backend was built with Node.js and Express.js, with MongoDB for data storage. I integrated Vapi for voice synthesis and recognition capabilities, allowing for realistic voice-based interactions. The AI personas were powered by OpenAI's API, with Langchain for conversation flow management and Pinecone for vector search to retrieve relevant responses based on the conversation context. The platform includes features for user management with role-based access control, a knowledge base for storing competitor information and training materials, persona management for creating and customizing AI personas, roleplay assignments for scheduling practice sessions, and comprehensive conversation tracking for performance review and improvement.