Building Low-Latency, Interruptible Voice Agents With Aws

What you'll learn
Build a virtual hotel receptionist voice agent with real-time reservation capabilities.
Build a fully functional low-latency voice agent using AWS Bedrock's Nova Sonic model.
Implement real-time HTTP/2 bidirectional audio streaming for instant AI responses.
Integrate Nova Sonic with DynamoDB to store and retrieve structured data.
Enable barge-in / interrupt support, allowing users to stop the AI mid-sentence.
Deploy a complete pipeline connecting audio input, Bedrock, tools, and UI visualization.
Capture microphone audio and stream it live to Bedrock using Python and PyAudio.
Understand Bedrock's bidirectional streaming protocol and event structure.
Build backend tools for identity verification, reservation queries, and booking modifications.
Handle asynchronous tool requests using Python's asyncio and background tasks.
Visualize streaming events, audio, and tool calls in Streamlit with a real-time dashboard.
Understand best practices for low-latency, human-like voice AI interactions.
Adapt the hotel use case to other industries such as, banks, customer success, etc..
Advanced Python Techniques
AWS Best Practices
AWS Quotas
AWS IAM role Best Practices
Requirements
Python Experience is needed
Internet + Laptop
Description
Build the next generation of real-time, human-like voice AI.In this hands-on course, you'll learn how to build low-latency, interruptible voice agents using AWS Bedrock's Nova Sonic, true HTTP/2 bidirectional streaming, and a fully integrated DynamoDB backend. This is not a toy project-this is a production-grade system with instant response times, barge-in (interrupt) support, real-time tool calling, and a complete event visualization dashboard.We will build a fully functional virtual hotel receptionist voice agent capable of verifying guests, checking reservations, reviewing balances, and updating bookings based on natural voice interactions. Every component is built from scratch using Python, asyncio, PyAudio, and Bedrock streaming APIs, giving you deep insight into how modern voice assistants actually work under the hood.You will also learn how to capture microphone input, stream it to AWS with minimal latency, play generated speech in real time, and gracefully handle interruptions mid-sentence. With DynamoDB, you'll build tools for identity verification and reservation updates-and connect them directly to Nova Sonic using Bedrock's tool calling system.To bring everything together, you'll build an interactive Streamlit dashboard that displays live events, audio chunks, tool calls, and model outputs-perfect for debugging, demos, or product development.By the end of this course, you will have a fully operational, real-time, interruptible voice AI pipeline that can be adapted to hospitality, healthcare, customer service, or any industry requiring human-like conversational systems
Machine Learning Engineers who want to build real-time, production-grade voice agents using AWS Bedrock.,AI Engineers looking to master bidirectional streaming, barge-in support, and human-like voice interactions.,Data Scientists who want to go beyond modeling and learn how to deploy live, low-latency voice systems.,Software Engineers interested in integrating LLMs, streaming protocols, and async Python workflows.,Cloud Developers & AWS Practitioners who want hands-on experience with Bedrock, DynamoDB, and serverless tooling.,Backend Engineers who want to build tool-calling pipelines and real-time database integrations.,Full-Stack Developers who want to incorporate advanced AI+audio features into real-world applications.,DevOps/Platform Engineers interested in supporting low-latency AI workloads in production environments.
Homepage