Reka develops enterprise-focused AI models that can be personalized to proprietary data and run on customer infrastructure or via Reka’s API. Its first commercial product, Yasa, is a multimodal assistant trained to understand images, videos, tabular data and text and is currently in closed beta. Reka also offers a service to adapt or “distill” large language models to a company’s datasets, with select customers able to run distilled models on their own infrastructure. The founding team includes researchers from DeepMind, Google, Baidu and Meta: Dani Yogatama, Cyprien de Masson d’Autume, Qi Liu Head and Yi Tay. Reka plans to expand Yasa to accept and generate more types of data and to build models that continuously self-improve without full retraining. The company is not yet generating revenue and will use its initial funding to acquire Nvidia compute and build a business team.

Total raised

$58M

Funding rounds

1

Latest round

Equity

Latest activity

Jun 2023