Rylo, the Israeli startup known until recently as Nagish, is expanding its communication technology for deaf and hard-of-hearing people through a new collaboration with Nvidia. According to CTech, the company will train its sign language recognition models on Nvidia's supercomputing systems and AI infrastructure, with the aim of translating sign language into text or speech in real time.
How it works
Translating sign language is especially hard for AI, because it means analyzing video: hand movements, facial expressions and body posture. According to reports, Rylo's research team uses Nvidia's Cosmos Tokenizer technology, which turns video into a compact digital representation that modern language models can process. The research is led by Dr. Amit Moryossef, founder of sign.mt, which Rylo acquired.
A small, profitable company
Rylo, founded by Tomer Aharoni and Alon Ezer, develops apps that provide real-time captions for phone calls. In June it raised $85 million at a valuation of about $500 million, led by General Catalyst and Canaan. According to the company, it is profitable and has only about 32 employees, about half of them in Israel. Its revenue comes mainly from a US Federal Communications Commission (FCC) license that lets it be paid from a federal program for every minute of captioned calls, rather than charging users. The company targets $100 million in annual revenue by the end of 2026.
Alongside the collaboration, Rylo also launched a free Mac app that generates live captions for meetings on Zoom, Google Meet and Microsoft Teams.
Why it matters
Nearly 50 million Americans live with hearing loss, as do hundreds of thousands of Israelis. AI that reliably translates sign language could change their ability to communicate with doctors, authorities and workplaces without waiting for an interpreter. It is also an example of Israeli high-tech building a profitable business around technology with a clear social impact.
