Substrate raises $8M to simplify AI application development

Substrate Labs Inc., operator of a cloud platform for building artificial intelligence applications, today announced it has secured $8 million in seed funding.

The funding was led by Lightspeed Venture Partners, with participation from South Park Commons, Craft Ventures, and several angel investors.

“New AI models hold great potential, but the tools available to developers today prevent them from introducing intelligence into software systems,” said Rob Cheung, co-founder and CEO of Substrate. “We built Substrate to make writing AI-integrated programs as intuitive as writing any other program: by connecting small semantic tasks together to automate work.”

Many AI applications rely on multiple neural networks rather than one to power their functionality. For example, a marketing automation tool might use one model to generate ad copy and a second to translate text into multiple languages. Substrate's namesake cloud platform is designed to simplify the development of such applications.

The company claims that its platform can host all neural networks used in a program in a single environment, which is simpler than the traditional approach of mixing neural networks hosted by several different AI providers. By eliminating the complexities associated with using multiple providers, Substrate promises to speed up software development.

The company's platform provides access to more than six prepackaged large-scale language models, including the Llama 3 model series from Meta Platform Inc. and several LLMs from well-funded AI developer Mistral. Substrate also offers neural networks optimized for other tasks, such as image generation.

In addition to AI models, the platform includes many other software building blocks often required for AI applications: a vector database, a database optimized for storing the mathematical structures that hold the information ingested by neural networks, and a built-in code interpreter that allows you to generate and run Python scripts with your application's AI models.

Substrate says its platform not only simplifies software development, but also improves the performance of AI workloads in the process. When an application's neural networks are hosted on different providers' platforms, moving data between them can take significant time. In contrast, Substrate says its platform can move data between AI models with latency measured in microseconds.

The company began developing its platform about a year ago and made it generally available today to coincide with the announcement of its seed round.

Image: Unsplash