Real-Time Voice AI Agent
Status:
Completed
A conversational, voice-driven AI agent supporting in-store customer service. It responds in real time and performs actions through tool calls, such as checking product availability or order status. I developed its administration panel, contributed to the agent itself and deployed it on Google Cloud.
Commercial project
Python
LangGraph
Pipecat
Gemini Live
Deepgram
React
GCP
Overview
The objective was to enable as many tasks as possible to be completed by voice, in an interaction that feels as natural as a conversation with a person. Built on real-time models (Gemini Live), the agent responds with minimal latency and invokes tools during the conversation, for example to check stock levels or the status of an order.
My role
- Developed the administration panel in React, enabling the team to add and configure tools, connect them to the agent and test its behaviour.
- Contributed to the development of the agent (LangGraph, Pipecat).
- Deployed the agent and provisioned its infrastructure on Google Cloud Platform.
Project Details
Project Type
Commercial · Voice AI
Timeline
2026