rjryan

Hi, I’m RJ.

I’m an AI researcher at Google DeepMind, where I work on multimodal LLMs, representation learning, and generative modeling.

I work on Gemini’s audio capabilities including realtime audio-visual dialog capabilities found in Gemini Live, Search Live, and Project Astra.

Every time you have a live conversation with Gemini, you’re using streaming infrastructure I created to talk to Gemini models I helped build.

Apart from my work on Gemini, I lead a team of 11 talented researchers in DeepMind’s Frontier AI organization. My research interests include:

• Multimodal language models, including Gemini.
• Speech understanding and generation.
• Full duplex and multi-stream language modeling.
• Joint representations of language in speech, text, images, and video.
• Generative modeling, self-supervised learning, and latent variable models.
• Streaming neural networks.

I’m passionate about learning and understanding the world around me. My preferred way of learning is by doing.

Elsewhere:

• X
• LinkedIn
• Scholar
• GitHub