/

Exclusive

Reka Launches to Develop Custom AI Models for Enterprise Applications

With its focus on customization and simplicity, Reka aims to empower enterprises to leverage AI models specifically designed for their unique needs and enhance their operational efficiency and insights.

by

San Francisco-based startup Reka has emerged from stealth mode with $58 million in funding from investors including DST Global Partners, Radical Ventures, and Snowflake Ventures. Founded by a team of researchers from DeepMind, Google, Baidu, and Meta, Reka aims to address the limitations of large language models (LLMs) by developing customized AI models for specific enterprise use cases.

While LLMs like OpenAI’s GPT-4 have gained popularity for their text analysis and generation capabilities, their generalist nature can pose challenges when organizations require precise instructions and consistency. Fine-tuning LLMs to narrow their scope is technically complex and costly. Recognizing the need for a simpler solution, Reka’s co-founders decided to develop models tailored to specific applications.

Reka’s first commercial product, called Yasa, is a multimodal AI assistant that goes beyond text to understand images, videos, tabular data, and language. Yasa can generate ideas, answer basic questions, and derive insights from internal data within an organization. What sets Yasa apart is its ability to be easily personalized to proprietary data and applications, allowing enterprises to benefit from the progress of LLMs without requiring extensive in-house AI expertise.

Reka plans to expand its offerings to include AI models that can handle various data types and continuously self-improve without the need for retraining. The startup also provides a service to adapt its pre-trained models to custom or proprietary datasets, enabling customers to run the distilled models on their infrastructure or via Reka’s API.

Reka faces competition from other startups like Writer, Contextual AI, LlamaIndex, and Cohere, which also focus on developing AI models suited for enterprise applications. Established players like OpenAI have also introduced tools for fine-tuning models and connecting them to external sources for ongoing updates.

Reka’s partnership with Snowflake, a strategic investor, allows Snowflake customers to deploy Yasa directly from their accounts. Additionally, Reka is collaborating with big data analytics company Appen to build tailored multimodal model-powered apps for enterprise clients.

With its focus on customization and simplicity, Reka aims to empower enterprises to leverage AI models specifically designed for their unique needs and enhance their operational efficiency and insights.

Related Stories