Neon

Conversational audio from real phone calls. Ethically sourced from consenting, paid speakers. For researchers who need data that reflects the real world.

[Neon @NeonAIResearch]

Engineering Lead: https://t.co/ywZNmI5UUr

Mobile Engineering Lead: https://t.co/ajF92r0vZZ

Senior Speech AI Researcher: https://t.co/dkC0JWto0w

[Neon @NeonAIResearch]

Frontier labs have a data problem they cannot scrape their way out of. Natural two-person conversation, consented and paid for, at scale.

Neon supplies it, and we are hiring. See links below.

[Neon @NeonAIResearch]

We're pleased to welcome Stan Kirdey, our newest AI researcher!

Stan is the founder of Clark, building AI labor systems. His prior experience includes: Head of Post-Training at Inflection AI, founding Amplifier Health, continual learning at Tenyx & NLP infrastructure at Netflix.

[Neon @NeonAIResearch]

Welcoming Dr. Rita Singh to Neon as an advisor!

Rita is a professor at @LTIatCMU and Director of the Center for Voice Intelligence & Security. She is an ISCA Distinguished Lecturer and speaker at @wef.

Her work focuses on using AI to extract insights from the human voice.

[Neon @NeonAIResearch]

Excited to welcome @sharmavasu55 as an advisor to Neon!

Vasu is Head of AI at PocketFM and was previously at Meta AI (FAIR), where he co-authored DINOv2 and Chameleon.

He has published 100+ papers with over 15,000 citations across NeurIPS, ICML, ICLR, CVPR, ACL & NAACL.

[Neon @NeonAIResearch]

Excited to announce Neon’s $25 million seed round led by Lightspeed with participation from Upper90, Upfront Ventures, and more awesome investors!

Neon is based on the crazy idea that people should get paid for their personal data. To our over half a million end users, we provide the ability to make hundreds or even thousands of dollars a year doing things they’d do anyway. To AI labs, we offer the ability to train models on ethically sourced studio-grade conversational audio and high-definition video. And to our current and future employees, we offer competitive salaries, equity, awesome colleagues, and the opportunity to swing for something really, really big.

And we're hiring! We’d love to hear from you if you’re a rockstar mobile developer, full stack engineer, data scientist, or voice/multimodal AI researcher! Jobs link here

https://t.co/y2pbbkkNfl

[Neon @NeonAIResearch]

Ever tried conversing with AI and noticed the speech seeming unnatural and awkward?

The reality of training AI on transcripts of text is like attempting to learn a dance from reading textbook instructions. Real human conversations are filled with flaws and non-verbal cues. To successfully bridge the current gap between AI agents that "speak" and AI agents that "understand", we can't treat voice data as mere text transcripts.

The future of voice AI is not just more synthetic generation but rather capturing the imperfect reality of human speech and connection.

[Neon @NeonAIResearch]

Voice is shaping up to be the natural interface for AI agents, just like talking to a personal assistant. We speak 4x faster than we type, and voice gives models the contextual awareness to fit seamlessly into our daily lives.

The bottleneck? High-quality, diverse conversational data barely exists. It's inherently private, there are no large open-source datasets, and that's historically slowed research in this space.

We're aiming to change that by releasing large-scale, real everyday conversations across diverse scenarios to bring us closer to a world where humans and AI agents work together seamlessly through voice.