Hoping to attract more enterprise teams into its ecosystem, Adobe Adobe launched a new model customization service called AI Foundry that will create custom versions of its flagship AI model, Firefly.
Adobe AI Foundry will work with enterprise customers to research and retrain Firefly Model Specific to the customer. AI Foundry Edition models are different from custom Firefly models because Foundry models understand multiple concepts compared to custom models that only have a single concept. These models will be should also be multimodalOffers a broader use case than the custom Firefly model, which can only ingest and react with images.
The Adobe AI Foundry model, based on Firefly, will learn the company’s brand tone, image and video style, products and services, and all of its IP. The models will generate content based on this information for any use case desired by the company.
Hannah Elsacre, vice president of GenAI New Business Ventures at Adobe, told VentureBeat that the idea to establish AI Foundry came about because enterprise customers wanted more sophisticated custom versions of Firefly. But given how complex the needs of enterprises are, Adobe will work to reengineer rather than hand the reins over to customers.
“We’ll retrain our Firefly commercially secure models with enterprise IP. We keep that IP separate. We never feed that back into the base model, and the enterprise itself owns that output,” Elsaker said.
Adobe will deploy the Foundry version of Firefly through its API solution, Firefly Services.
Elsakr compared AI Foundry to a consulting service, as Adobe will have teams working directly with enterprise customers to retrain models.
deep tuning
Elsaker refers to Foundry as a deep tuning method because it goes far beyond simply fine-tuning a model.
“The way we think about it, probably in layman’s terms, is that we are surgically reshaping the firefly-based model,” Elsaker said. “So you get the benefit of all the knowledge in the world from our image model or video model. We’re going back in time and bringing in IP from an enterprise like a brand. It could be footage from a one-shot genre, whatever they have the license to contribute. We then re-train. We call it continuous pre-training, where we overweight the model to dial in certain things differently. So we literally re-train our base model. are trained, and that’s why we call it deep tuning instead of fine-tuning.”
One part of the training pipeline involves Adobe’s embedded teams working with the company to identify the data they’ll need. The data is then securely transferred and captured before being tagged. This is fed into the base model, and then Adobe starts a pre-training model run.
Elsacher says the Foundry versions of the Firefly will not be smaller or distilled models. Often, additional data from companies expand the parameters of Firefly.
Two early customers of Adobe AI Foundry are The Home Depot and Walt Disney Imagineering, Disney’s research and development arm for its theme parks.
“We’re always looking for new ways to enhance our customer experience and streamline our creative workflows. Adobe’s AI Foundry represents an exciting step forward in adopting cutting-edge technologies to deepen customer engagement and deliver impactful content across our digital channels,” said Molly Battin, senior vice president and chief marketing officer of The Home Depot.
more customization
Enterprises often turn to Fine-tuning and model optimization Bringing large language models closer to your company’s needs with our vast external knowledge. Fine-tuning enables enterprise users to use the model only in the context of their organization’s data, so the model doesn’t respond with text completely unrelated to the business.
However, most organizations make improvements on their own. They connect to the API of the model and start retraining it to give answers based on their ground truth or their preferences. Several methods exist for fine-tuning, including some that can be done with just a hintOther model providers also try to make it easier for their customers to fine-tune models, e.g. OpenAI with o4-mini reasoning model,
Elsacre said he expects some companies to have three versions of Firefly: a Foundry version for most projects, a custom Firefly for specific single-concept use cases, and base Firefly because some teams want a model less burdened by corporate knowledge.
