Based on the premise that it is impractical to deploy an all-purpose LLM for specific use cases, a group of researchers from Google, Baidu, DeepMind and Meta founded Reka AI in July 2022. A year later the company has emerged from stealth mode with news of $58 million in Series A funding led by DST Global and Radical Ventures. Strategic partner Snowflake Ventures also participated, along with angel investor Nat Friedman, former CEO of GitHub. The Sunnyvale, California-based startup says it is building “enterprise-grade state-of-the-art AI assistants for everyone, regardless of language and culture.”

Reka CEO and co-founder Dani Yogatama told TechCrunch that the company’s approach “allows enterprises to benefit from progress in LLMs in a way that satisfies their deployment constraints without requiring a team of in-house expert AI engineers.”

At the moment, Reka is also offering select clients the option of fine-tuning an LLM on their proprietary data, a service it plans to expand. “Customers can run the ‘distilled’ models on their own infrastructure or via Reka’s API, depending on the application and project constraints,” TechCrunch reports.

Reka’s first product, Yasa, is still in closed beta. The multimodal “assistant” does more than generate text; it also understands photographs, videos and tabular data.

“It can be used to generate ideas and answer basic questions, Yogatama says, as well as derive insights from a company’s internal data,” explains TechCrunch, adding that while similar to GPT-4 in those basic parameters “the twist is that Yasa can be easily personalized to proprietary data and applications.”

Reka plans to use its new funds to build generative models and conduct AI research in universal intelligence, specifically “general-purpose multimodal and multilingual agents, self-improving AI and model efficiency.” On the drawing board is AI that can ingest and generate a wider range of datasets. Another focus is improving autonomous learning, so less human intervention will be required to keep the system up to date.

Other AI startups working on models targeting enterprise use cases include San Francisco-based Writer, geared to editorial enterprises, and Contextual AI, which is developing contextual foundation models out of New York. Another, LlamaIndex, has created an open source framework that feeds custom data into LLMs.

The market for custom LLMs heated up considerably with Databricks’ recent $1.3 billion deal to buy model training company MosaicML.

Reka investor Snowflake “recognized the potential of Reka’s specialization in developing large AI models for enterprise clients” and announced an initiative allowing its customers “to run third-party model providers like Reka within their Snowflake account,” reports InfotechLead.

Snowflake also unveiled “a collaboration with Nvidia, granting enterprises access to Nvidia’s software stack for AI model development,” InfotechLead says.

Related:
Nvidia’s NeMo Delivers AI Customization to Snowflake Cloud, ETCentric, 6/30/23