Today, every company is investing in AI initiatives after the explosion of generative AI in the past six months. Those that don’t will get left behind. Yet many companies don’t know how to implement AI into their business operations. Customizing models is extremely difficult, requiring expertise in ML Infrastructure and experience with a large number of software packages. Often these software packages are cobbled together and don’t achieve the desired results. Enterprises rarely have the internal expertise or time to ensure AI is done right.

So today, we’re launching an open-source platform for customizing large language models (LLMs) for enterprise use cases. The platform allows enterprises to customize open-source LLMs to achieve optimal performance for their specific use cases using their own data. Businesses can more easily build applications for their highest-priority generative AI use cases. Fine-tuning open-source models with Scale will give organizations better performance for their specific use cases, increased customizability, and enhanced safety at a lower total cost than other solutions.

As part of our efforts, Scale is proud to be a Meta Llama 2 Launch Partner, and we're excited to make it easier to customize open-source LLMs like Llama 2. Scale CEO and founder Alexandr Wang also signed a statement in support of Meta’s open approach to AI that enables everyone to benefit from the technology and brings additional levels of visibility, scrutiny, and trust.

What Scale is Making Possible

With this new offering, enterprises can customize open-source models with just a few lines of code leveraging their own data. For more demanding model customization needs, Scale’s fine-tuning platform, combined with our Data Engine, will optimize model performance for the most challenging business use cases. The Scale Data Engine powers the most advanced LLMs and generative models through world-class RLHF, data generation, model evaluation, safety, and alignment, making it easy for every enterprise to adopt industry-leading generative AI.

Scale helps insurance companies summarize and classify claims more efficiently, retail and e-commerce companies improve the shopping experience through personalized recommendations, and wealth managers increase returns by quickly summarizing and identifying the most relevant trends from financial data. Achieving this performance at these specific tasks is not possible without highly capable base models that are fine-tuned with high-quality data.

Our new platform enables:

Performance: Models customized with proprietary data perform better at high-priority tasks.
Customizability: Customize models to specific use cases, and shape how they respond to common scenarios so they can be an effective extension of the brand.
Safety: Enhance the safety of models with human-in-the-loop testing, evaluation, and monitoring.

How to Get Started

To support an open innovation approach to AI, Scale is providing an open-source package and open APIs for model fine-tuning and hosting, along with clear documentation on how to use these APIs. The package and APIs empower any organization, no matter their expertise, to get up and running quickly and easily. Enterprises that need an additional performance boost for their specific use cases can leverage the power of human experts to enhance their data with the Scale Data Engine. Ultimately, this platform will deliver fine-tuned open-source models for organizations’ most important use cases and for specific domains to encourage the adoption of generative AI across industries.

Get started by downloading the open-source repo from Github, or if you are looking to optimize your models with our enterprise model customization platform and the power of the Scale Data Engine, contact us today.