Amazon Web Services

Innovation requires more than just ambition: It can demand scale, efficiency, and access to massive computational power. BRIA AI—which offers a developer-focused solution to build visual foundation models—trains models from scratch using fully licensed, legally sourced commercial data, so it needs the ability to process large amounts of data for images, video, and audio. As demand for the company’s high-quality Generative AI solutions increased, BRIA AI decided to work alongside Amazon Web Services (AWS) and AWS Partner Automat-it to build a scalable, cost-effective architecture for model training and data management. Now, BRIA AI can focus on innovation while maintaining operational efficiency, cost effectiveness, and developer-first accessibility.

Opportunity | Keeping Engineering Teams Focused on Developing Models—Without Worrying About Infrastructure

As BRIA AI expanded, managing large amounts of data and scaling complex training jobs became more and more difficult. The company needed a trusted collaborator to provide technical expertise and flexible infrastructure support so that its engineering teams could stay focused on what they do best: developing open, commercially viable generative AI models.

To help support its teams, BRIA AI chose to work with Automat-it and AWS. The company had worked alongside AWS before, so BRIA AI trusted the cloud provider’s ability to deliver secure, high-performance infrastructure. And working with Automat-it brought the operational expertise that helped BRIA AI to shift from limited in-house GPU clusters to a powerful, purpose-built solution capable of meeting enterprise-level demands.

Solution | Building a Supercomputing Environment to Support Innovation

Working closely with BRIA AI, Automat-it helped design, provision, and optimize a supercomputer architecture on Amazon SageMaker, which delivers an integrated experience for analytics and AI with unified access to all a company’s data. Automat-it also facilitated seamless integration of AWS Glue and Apache Iceberg for managing data lakes, alongside intelligent data storage using Amazon Simple Storage Service (Amazon S3)—object storage built to retrieve any amount of data—along with Intelligent-Tiering and Glacier for cost efficiency.

Using this solution, BRIA AI could scale its architecture effortlessly—something that previously required considerable effort to manage. Automat-it took on the heavy lifting of infrastructure management and cloud architecture optimization, giving BRIA AI the freedom to focus entirely on model innovation and development. “Four years ago when BRIA AI started, we began working very small on few GPUs and a little amount of data,” says Bar Fingerman, vice president of engineering at BRIA AI. “Fast-forward to today, we’re using supercomputers on Amazon SageMaker, streaming petabytes of data to the cloud and to the cluster, and training foundation models for weeks instead of months.”

Outcome | Shaping the Future of Responsible Generative AI

Through its collaboration with Automat-it and AWS, BRIA AI has achieved the scale needed to meet rising industry demands while staying laser-focused on its core competencies. By offloading infrastructure management to Automat-it, the company can now dedicate its engineering talent to enhancing its developer-centric solution and creating high-impact, commercially licensed foundation models for the visual domain.

Today, BRIA AI is operating at a level that was once attainable only by the most well-funded technology firms. The company’s open-source and developer-first approach is shaping how responsible generative AI can be developed commercially. Looking ahead, BRIA AI will continue to work with Automat-it to onboard new capabilities and push the boundaries of what’s possible in generative AI, all while keeping developers focused on innovation.

customer-stories
generative-ai-trailblazers
ai

Up Next

VideoThumbnail
11:09

Inside Amazon's High-Tech Fulfillment Centers: How AWS and Automation Deliver Prime Packages in One Day

Nov 22, 2024
VideoThumbnail
48:52

Modernizing Nintendo eShop: From Monolith to Microservices - A Platform Engineering Journey on AWS

Nov 22, 2024
VideoThumbnail
1:31:17

AWS re:Invent 2023 - Partner Keynote: Achieving Impossible Firsts with AWS and Strategic Partnerships

Nov 22, 2024
VideoThumbnail
57:25

How Electronic Arts Transformed Its Data Platform: Migrating to Amazon EMR for Improved Performance and Cost Savings

Nov 22, 2024
VideoThumbnail
3:18

NinjaTech AI's Revolutionary Multi-Agent Personal Assistant: Powered by AWS Trainium and Inferentia

Nov 22, 2024