OneAdvanced, a UK-based enterprise software provider, deployed over 50 AI agents on a UK-sovereign AWS architecture to meet data residency and compliance requirements. The solution leverages Llama 4 Maverick and Llama Guard 4, hosted on Amazon SageMaker AI, to deliver AI capabilities while ensuring no data leaves the UK. The architecture includes a Retrieval Augmented Generation (RAG) pipeline, over 50 Strands agents, and a tool layer on Amazon ECS. This approach supports ISO 42001 certification for AI governance and full control over model serving infrastructure. The solution enables rapid agent deployment and maintains strict data residency standards for customers in healthcare, legal, and other regulated sectors. Source: awsml

OneAdvanced initially used Amazon Bedrock for prototyping but switched to self-hosting due to data sovereignty requirements. The models Llama 4 Maverick and Llama Guard 4 were not available through AWS managed services in the UK at the time. To comply with regulations, OneAdvanced deployed these models on their own AWS accounts, using vLLM on p5.48xlarge instances in the London region. The solution includes content moderation, document retrieval, and a no-code agent builder for non-technical users. Load testing with vLLM validated the infrastructure could handle throughput requirements, and OneAdvanced moved from p4d.24xlarge to p5.48xlarge instances for production, taking advantage of reserved instance discounts. Source: awsml

Andrew Henderson, CTO of OneAdvanced, emphasized the importance of data sovereignty for customers in the public sector and regulated industries. He stated, "Data sovereignty, particularly in the UK, is a hard requirement for many of our customers, especially those in the public sector and highly regulated industries. They need to know exactly where their data is, who has access to it, and that it resides within the UK’s legal and regulatory framework to support total compliance and trust." The solution also includes Llama Guard 4, which checks user inputs for harmful content before the main model processes the request. OneAdvanced built over 50 agents in three weeks, with most developed in under a day, using the Strands Agents SDK. Source: awsml