Israeli Startup Rylo Partners With Nvidia to Advance AI Sign Language Translation

Israeli startup Rylo, which develops technology for the deaf and hard of hearing, has signed a cooperation agreement with Nvidia. This partnership comes as Rylo experiences significant growth, having recently secured $85 million in funding after obtaining a license from the U.S. Federal Communications Commission (FCC) for its real-time AI-powered text-to-speech and speech-to-text technology. The collaboration with Nvidia will enable Rylo to accelerate the development of its additional product: real-time AI-based sign language translation. Rylo acquired sign.mt, a company led by Dr. Amit Mor-Yosef, about a year ago. The startup plans to leverage Nvidia's models to train its sign language translation solution, which utilizes Nvidia's Cosmos Tokenizer model. This model integrates naturally with modern language models like GPT and facilitates video generation and reconstruction, forming a basis for research in understanding and translating sign language in real-time. Nvidia's infrastructure, including Megatron Bridge and NeMo Framework, will be used for training these models. The current market for transcription services and sign language interpreters is estimated at $4 billion annually, with interpreters costing around $75 per hour. Rylo states that existing tools only serve 5% of this market due to a shortage of sign language interpreters. "Training models capable of reliably interpreting the nuances of sign language and converting them to text or voice in real-time requires significant computing resources. Nvidia's software and hardware infrastructure allows us to train the models at a broader and more accurate scale, and at a much faster pace than was previously possible," said Mor-Yosef. Founded in 2022 by Tomer Aharoni (CEO) and Alon Azar, Rylo was initially named NAGISH but rebranded to align with its expansion into the American market. The company's application is designed for the deaf and hard-of-hearing community to facilitate phone conversations, an area where traditionally users rely on specialized call centers. Rylo's app transcribes entire conversations, allowing users to type or speak. Typed input can be converted to real-time speech for the other party. Rylo is one of only five companies to receive the FCC license, and notably, it is the only one not reliant on a human call center, operating entirely through technological means. The FCC license is crucial as it means the commission, not the users, pays for the service, with fees ranging from $1 to $8 per minute of conversation, reviewed every three years. Since receiving the license, Rylo has begun generating initial revenue and projects $100 million in revenue by the end of 2026. The company is already profitable and aims to reach an annual revenue run rate of $1 billion by 2028, incorporating its sign language translation product line. To date, Rylo has raised $101 million from investors including General Catalyst, Canaan, Precursor, K5, Vertex, and Kardan.