Your Voice Agent is Just a Walkie-Talkie
Neil Zeghidour
- When
- Tuesday, June 3012:05 PM – 12:25 PM · 20 min
- Where
- Track 6San Francisco, CA · imported from ai.engineer's public schedule feed
About this session
Everyone says cascaded voice pipelines are dead and native speech models are the future. Yet production environments are still dominated by STT-LLM-TTS stacks. Reconciling the natural flow of native audio with the elite reasoning of a cascaded agent remains an unsolved systems problem. This talk dissects the brutal technical trade-offs behind that counterintuitive reality. We will break down why your voice agent is still stuck behaving like a walkie-talkie and map out the specific technical roadmap required to build full-duplex AI that actually works.
Speaker
Co-founder & CEO, Gradium
Neil Zeghidour is the co-founder and CEO of Gradium. Neil founded Gradium after a decade of building and leading frontier generative audio teams at Meta and Google DeepMind. Being frustrated by slow and brittle voice assistants , he built the engineering teams that developed the first neural audio codecs and introduced the first audio LLMs, such as AudioLM, at Google. He later created Kyutai to launch Moshi, the world's first real-time, full-duplex conversational AI , and Hibiki, the first simultaneous speech-to-speech translation system. Today, Gradium is focused on helping developers build natural, real-time voice agents by providing ultra-low latency streaming APIs that transition these breakthroughs from the research lab to production.
More in Claws & Personal Agents
- Security Firewall for AgentsTuesday, June 30 · 10:45 AM – 11:05 AM · Track 1
- Governance Is the Real Bottleneck to AI ROITuesday, June 30 · 10:45 AM – 11:05 AM · Leadership 2
- Your Agent Didn’t Fail. Your Harness Did.Tuesday, June 30 · 11:10 AM – 11:30 AM · Track 1
- Everyone Gets A Software CompanyTuesday, June 30 · 11:40 AM – 12:00 PM · Track 1
- Tethered: Our Agents Are UsTuesday, June 30 · 12:05 PM – 12:25 PM · Track 1
For developers: this programme is open data — JSON, iCal, schedule XML and an MCP endpoint.Show endpointsHide
- JSONEvery published session and speaker, in one request./aie-worldsfair-2026-import/feed.json
- iCalSubscribe in Google, Apple or Outlook Calendar./aie-worldsfair-2026-import/feed.ics
- Schedule XMLfrab / pentabarf — the format conference apps import./aie-worldsfair-2026-import/feed.xml
- MCP + RESTPoint Claude at the programme. OpenAPI 3.1 included./agents
No key, no signup, CORS open. Everything here is generated from the same data the organisers edit.