Building AI voice agents is no longer just about connecting an LLM with speech-to-text and text-to-speech.

Realtime Voice models give far more superior results.

However, there also break for certain use cases.

In this talk, we'll explore practical patterns for building reliable AI voice agents using realtime voice models.

We'll cover challenges such as:
- Handling simultaneous voice and UI interactions
- Reducing latency while the agent is reasoning
- Keeping application state synchronised between the UI and the voice agent

The audience will leave with practical engineering patterns that make AI voice agents feel fast and, more importantly natural