Talk to MEmo
Team consisting of Memo CTO Jeroen Coupe and Software Engineer Dylan Garcia Calero, building conversational AI applications for telephony.
Project Description
Project Description
Core Functionality, Stability & Complexity Talk To MEmo is a multimodal social training agent designed to cure “approach anxiety” through realistic voice simulations. The system features a sophisticated state machine that orchestrates Twilio voice calls into a dynamic roleplay environment. It utilizes advanced tool orchestration to transition data seamlessly from real-time audio streams to a structured database, generating “David Attenborough-style” voice introducton and detailed CRM-style analytics accessible via a web dashboard. The prototype demonstrates high stability in handling the full lifecycle: audio ingestion, latency-optimized dialogue generation, sentiment analysis, and multi-channel delivery (Voice to SMS to Web).
Innovation, Creativity & Real-World Impact In an era of “flirt paralysis” and “hesidating,” Talk To MEmo innovates by shifting the industry focus from finding matches to building the skills to keep them. Unlike dating apps that gamify rejection, this solution gamifies self-improvement.
Theme Alignment
Our project perfectly embodies the theme by transforming disparate inputs—Voices (real-time telephony), Clouds (serverless inference), and Tools (CRM analytics)—into a single, cohesive agent that travels with the user.
Browsers: The agent pushes data to a reactive web dashboard for deep reflection.
Voices: The agent lives on the phone line, utilizing distinct voice synthesis for characters and narrators.
Technology Stack
Telephony & Messaging: Twilio Programmable Voice & SMS API.
LLM Orchestration: OpenAI GPT-4o (for dynamic dialogue generation and JSON-structured evaluations).
Voice Synthesis (TTS): ElevenLabs (for high-fidelity character voices and the “Attenborough” narrator profile).
Speech-to-Text (STT): OpenAI Whisper or Deepgram (for low-latency transcription).
Backend Framework: Node.js.
Database: Neondb (storing CONVERSATIONS, EVALUATIONS, and USER_STATS).
Frontend: React(for the CRM-style Dashboard and annotated transcripts).
Prior Work
https://libraries.io/npm/create-twilio-agent