Typeface expands customized generative AI approach with Google Cloud partnership

Typeface expands customized generative AI approach with Google Cloud partnership

Join top executives in San Francisco on July 11-12, to hear how leaders are integrating and optimizing AI investments for success. Learn More

Typeface is continuing to advance its customized generative AI agenda with a new Google Cloud partnership announced today.

The San Francisco-based company emerged from stealth in February with $65 million in funding to help build out customized AI technology for enterprises looking to generate marketing and branding content. The startup is led by former Adobe CTO Abhay Parasnis, and its goal is to help bring the power of generative AI to big brands across multiple industries that don’t get what they need from generalized large language models (LLMs).

The Typeface platform enables organizations to train LLMs for a specific brand or use case to get customized results for both images and text. The Google Cloud partnership will see the latest LLMs from Google — including those based on PaLM 2 LLM — integrated into Typeface. 

Going a step further, Google and Typeface have a go-to-market partnership through which Typeface’s customized AI technology can be directly integrated into Google Workspace.


Transform 2023

Join us in San Francisco on July 11-12, where top executives will share how they have integrated and optimized AI investments for success and avoided common pitfalls.

Register Now

“This is what we call affinity AI,” Parasnis told VentureBeat. “This is the next stage of generative AI, where companies will take generic models but customize them uniquely to their products, their voice, their customer and their audiences.” 

Inside the Typeface AI platform, it’s more than a zero shot model

Customizing existing LLMs is commonly achieved with a zero shot approach to fine tuning that doesn’t require much (if any) additional training. But, the approach doesn’t always yield the best results.

Parasnis explained that Typeface is going beyond zero shot to help to build out customized models based on existing LLMs. The training is company-specific and aims to address any potential privacy concerns.

“We will give a company a proprietary container of AI models that they own and control and it doesn’t flow back into the broader model,” said Parasnis. “If you’re a big brand, your content is an asset and you don’t want it getting used or misused by many other people.”

A layer called the Typeface Graph — a proprietary technology that can sit on top of existing LLMs — is like a data lake, but it is multi-model as it understands images, text and videos to help create a  rich metadata model of an organization’s content, Parasnis explained. On top of the Typeface Graph is a vector database which can then help with data retrieval and interfacing with the LLMs.

TypeFace FLOW is like LangChain, but for enterprises

Simply generating a piece of text or an image is often only one piece of an organization’s workflow. For example, an organization might want to generate an image and text for a marketing campaign.

In the developer community, the open source LangChain tool is increasingly being used to chain multiple generative AI prompts and models. Parasnis noted that Typeface is using LangChain internally, but it is a developer tool. Typeface’s Flow service is doing the same thing LangChain does at the developer level, but doing it for higher level business workflows for generative AI.

“Flow is more for business users to say, ‘I want to do Instagram posts, or I want to do a Google ad,'” Parasnis said. “That’s not what a LangChain user can do.”

Google already integrates generative AI, Typeface goes a step further

As part of the Google partnership, organizations will be able to directly integrate Typeface with Google Workspace applications. Parasnis said this differs from what Google itself is already doing.

At its I/O event in May, Google announced its own Duet Generative AI services for Google Workspace. Parasnis said that Google is doing a lot of work in their own applications to integrate generative AI, although in his view that work is more generic. As Typeface is trained on an organization’s specific data, Parasnis said it can provide a level of deep customization that a general model cannot achieve.

“Think of [Google’s efforts] as much more horizontal innovation for everyone who uses Workspace,” he said. “We are focusing on specific enterprise use cases.”

Google and Typeface are hardly strangers, either. Parasnis said that Google Ventures is an investor in the company, as is Microsoft’s Venture Fund M2. Parasnis also said his company has a partnership with Microsoft.

“Our intention here is to establish Typeface as the preferred enterprise generated platform that works with many enterprise companies including Microsoft, Google and others, ” he said.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.