Seattle-based startup Fixie.ai Inc. today said that it has raised $17 million in seed funding to help companies integrate large language model AI agents into their applications.

Large language models, or LLMs, are a type of artificial intelligence tool that can read, summarize and generate text or other content and even hold conversations. Its use has become extremely popular of late with the emergence of chatbots such as ChatGPT from OpenAI LP, which has been integrated into Microsoft Corp.’s products such as its search engine Bing.

Fixie provides a cloud-based platform that allows developers to build, host and scale up AI agents that can integrate directly into their applications using any tools, data sources and application programming interfaces available. The vision of the company is to provide them with everything they need to use any LLM they choose.

“Fixie represents a foundational shift in building software, using LLMs as a computational engine,” Chief Executive Matt Welsh said in a blog post. “With Fixie, you wrap an existing data source or API in a little bit of code that hooks it up to a large language model of your choice. The resulting agent can then take English queries from end users, or even from other agents, and translate them into calls to that external system.”

Welsh is a former engineering leader at machine learning startup OctoML Inc. and before that was a principal engineer at Google LLC. His co-founders include Chief Technology Officer Justin Uberti, a former veteran Google engineering leader, as well as Chief Product Officer Zach Koch from Shopify Inc. and Google, and Chief AI Officer Hessam Bagherinezhad, who had worked with Welsh at Xnor.ai Inc. before Apple Inc. acquired it in 2020.

Fixie is launching its platform as a developer preview, allowing anyone to integrate an LLM into their own application, data or tool. It arrives shortly after OpenAI released its own plugin for ChatGPT that permits developers to extend the capabilities of the company’s GPT-4 powered chatbot.

The company’s platform supports GPT-4 out of the box, but it’s designed to allow companies to connect to whatever LLM they want. Fixie also supports commercial and open-source models such as AI21’s Jurassic-2, GPT-NeoX and Cohere. The platform handles all the underlying LLM interactions and all details for developers, such as user identity, authentication, session management, storage and configuration.

There’s a wide variety of use cases for Fixie, such as for customer support, business automation, business intelligence, content generation and customer relationship management automation. For example, a customer could use the platform to read customer ticket information and answer English queries about order history, produce information for a support agent when making a return label or answer customer questions.

With Fixie, LLM technology can be easily and directly integrated into its application at whatever level it desires and customize how it operates. Fixie delivers a core platform that connects to the data sources and APIs, then lets developers choose how they want those to be represented in their apps.

The seed round was led by Redpoint Ventures, with participation from Madrona Venture Group, Zetta Venture Partners, SignalFire, Bloomberg Beta, and Kearny Jackson. Welsh said the new funding would allow the company to build out the team and expand the reach of the platform.

“Existing LLMs are incredibly powerful and are only going to get more advanced, but the true magic comes from combining them with internal data and systems to dramatically improve the way businesses operate,” said Erica Brescia, a managing director at Redpoint. “That’s precisely what Fixie delivers.”

Access to the Fixie Developer Preview is currently available through the company’s API portal and developers can get access to the open-source software development kit and example code via the GitHub repository.

Image: Fixie