← All speakers

Bio, Work & Ideas

Yaxin

Yaxin participated in an AI Engineer World's Fair 2025 session. The session covered real-time voice architecture, conversational response latency, and streaming audio transport.

Read the topics behind these talks

1 conference talk

Key ideas

Scroll to read ↓

From conference Wi-Fi to a talking Raspberry Pi, natural voice interaction depends on a tight latency budget, the right media transport, and room to experiment.

  • An anxious stuffed animal has a point
    0:22 ↗
  • Measure the gap the listener experiences
    1:47 ↗
  • Choose the transport for its job
    3:25 ↗
  • Reliable delivery can make audio arrive too late
    4:57 ↗
  • Let the media stack handle the media
    6:42 ↗
  • An established transport for an unsettled interface
    7:45 ↗
  • Squabbert's local stack and uneven poetry
    9:59 ↗
  • Keep the device, change where the intelligence runs
    12:23 ↗
  • A first prototype for bilingual families
    13:09 ↗
  • Make the next experiment easier to build
    15:13 ↗

References