Ideas beneath
the surface

As enterprise usage of AI grows, and as AI models proliferate for both general and specific tasks, choosing the right model will become increasingly important. But, choosing the right model is only part of the challenge; prompting an AI model to elicit the desired output is difficult, manual in nature, and critical to driving model utility.

That’s why we invested in Not Diamond via Deepwater Venture Fund II.

Last year, Not Diamond created the world’s most powerful AI model router. By combining multiple LLMs into a “meta-model” that learns when to route a given query to a certain LLM, the company’s solution can outperform every individual model by routing effectively between them. More recently, Not Diamond released groundbreaking technology around model prompting, known as Prompt Adaptation. Prompt Adaptation automatically adapts prompts across models, replacing manual prompt engineering resources and reducing compute requirements.

In theory, any company (or individual) using AI models could be a customer of Not Diamond; as such, the addressable market is largely uncapped and difficult to quantify. As the uses and range of options for AI models continue to evolve, we expect increasing choice and uncertainty around which model(s) will best suit specific use cases and how best to prompt those models for high utility output. Today, Not Diamond offers self-serve volume-based pricing and enterprise packages. The product runs securely on top of customer data, continuously improving its prompts and model recommendations as teams scale. This creates a sticky data flywheel over the increasingly commoditized model layer.

SAP recently announced that it is leveraging the Not Diamond Prompt Adaptation breakthrough within its own Prompt Optimization Service. This SAP offering is intended to drive AI model accuracy improvements and accelerate engineering throughput. But SAP is not just an integration partner, it is a direct customer, using Not Diamond to radically accelerate engineering efficiency while improving output quality for internal SAP teams. We see SAP’s internal usage and integration of Not Diamond’s technology as a strong early validation of product-market fit.

Deepwater invested in Not Diamond’s most recent funding round alongside IBM, SAP.io Fund, defy.vc, Myriad Venture Partners, and others. Pre-existing investors include many of the world’s leading AI scientists, engineers, and executives, such as Jeff Dean (Google Chief Scientist), Julien Chaumond (co-founder & CTO Hugging Face), Zack Kass (OpenAI), Ion Stoica (Anyscale, Databricks), Tom Preston-Werner (GitHub), Scott Belsky (Adobe), and Jeff Weiner (LinkedIn).

Disclaimer