Beyond Text: Future of Voice AI
Join us for an evening of research talks with the people actually building voice AI — writing the papers, training the models, solving the problems the rest of the industry is still naming.
Speakers
Arjun Jain — Chief Scientist, Smallest.ai
PhD from Max Planck Institute, worked with Turing Award winner Yann LeCun at NYU, Adjunct Faculty at IISc. His research spans generative AI, computer vision, and speech — with a long view on how multimodal architectures evolve from research to production.
Talk: The Speech-to-Speech Landscape - Half-Duplex, Full-Duplex, and What It All Means. A researcher's worldview on where S2S models are today, how half and full-duplex architectures compare, and what the design choices actually imply for real conversational AI.
Sameer Khurana — Lead AI Researcher, Smallest.ai
PhD from MIT CSAIL (Spoken Language Systems group), followed by research on neural audio codecs and speech translation at Mitsubishi Electric Research Labs and Ultravox AI. His work includes foundational research on hierarchical audio representations and unsupervised speech learning.
Talk: From Audio Language Models to Conversational World Models. How to build the next generation of voice agents from first principles, and what the research actually says is possible today.
Event schedule
5:00–5:30 — Arrival + drinks
5:30–6:15 — (30 mins talk + 15 mins Q&A) From Audio Language Models to Conversational World Models by Sameer Khurana
6:15–7:00 — (30 mins talk + 15 mins Q&A) The Speech-to-Speech Landscape by Arjun Jain
7:00–8:00 — Networking with speakers and attendees
Who should attend
Engineers and product leaders building on or evaluating voice AI infrastructure.
AI researchers and applied scientists working on speech, audio, or language systems — whether at a lab, a startup, or a university.
Founders and technical operators who want the unfiltered view on what's production-ready versus what's still a research problem.
Investors tracking the voice AI stack and looking to understand where the defensible technical moats actually are.
Hosted by
Smallest.ai is an AI research lab building frontier speech-to-speech systems with async architectures designed for the latency and accuracy constraints that production deployments actually expose.
Notes
Capacity is limited. We’ll be recording and streaming the session for recap and promotion. By attending, you consent to being photographed/recorded and to use of those assets by the organizers.