JD

Dialex AI Phone Caller

Featured

A full-stack project management application that streamlines client communication, project tracking, and invoice generation for independent contractors.

November 15, 2024
PythonTwilioOpen AIMulti ThreadingWeb SocketsDocker
Dialex AI Phone Caller

Overview

Dialex is an AI-powered outbound calling system designed to handle real conversations at scale with near-instant response times. Built on Twilio, OpenAI, and a custom high-speed text-to-speech engine running in a self-hosted Docker container, Dialex can carry out both scripted and AI-driven calls. At its core is a modular, node-based system where each node represents either an AI response or a predefined script with dynamic placeholders. This setup allows users to build complex, branching call flows that combine deterministic functions such as cosine similarity with generative AI reasoning, all while running in parallel to achieve sub-second latency.

My Role and Contributions

  • AI and Voice Architecture: Designed a modular node system that blends scripted logic with AI nodes, giving full flexibility for building unique call structures.
  • Low-Latency Engineering: Built multi-threaded pipelines and optimized WebSocket handling to push response times below one second.
  • Custom TTS Integration: Developed a self-hosted, fast text-to-speech library in Docker for high-quality, real-time voice synthesis.
  • Function Routing: Implemented intelligent branching with functions like cosine similarity checks and LLM determinism to improve contextual accuracy during calls.

Recognition and Results

My Role and Contributions

  • Achieved real-time AI conversations with natural latency under one second, solving a major challenge in AI voice systems.
  • Demonstrated scalable outbound calling workflows that can switch between AI-generated and scripted dialogue seamlessly.
  • Proved viability for use cases including customer support, sales outreach, and automated surveys.

Why This Matters

Dialex shows my ability to integrate AI, telephony, and systems engineering into a cohesive product that addresses a complex real-world challenge: making AI conversations sound natural. It highlights my strengths in modular system design, latency optimization, and production-ready AI integration. These skills directly translate to building advanced, AI-native web and voice applications for clients.