OneAdvanced deploys over 50 AI agents on UK-sovereign AWS
Meta
OneAdvanced, a UK-based enterprise software provider, deployed over 50 AI agents on a UK-sovereign AWS architecture, self-hosting Llama 4 Maverick and Llama Guard 4 on Amazon SageMaker AI to ensure data residency. The solution uses a RAG pipeline with pgvector, the Strands Agents SDK, and runs on Amazon ECS.
OneAdvanced, a UK-based enterprise software provider serving over 10,000 customers, needed to deliver AI capabilities while ensuring no data leaves the UK. Since the specific models they wanted, Llama 4 Maverick and Llama Guard 4, were not yet available through AWS managed services in the UK region, they self-hosted open-weight LLMs on AWS infrastructure they fully control. The architecture includes Llama 4 Maverick and Llama Guard 4 served via vLLM on Amazon SageMaker AI endpoints, running on p5.48xlarge instances in the London (eu-west-2) region. Over 50 Strands agents run on Amazon ECS, each with its own system prompt and tool configuration, with agent configuration stored in Amazon DynamoDB. Documents uploaded to Amazon S3 are converted to markdown, chunked, and embedded into pgvector for retrieval. Llama Guard 4 checks user inputs for harmful content before the request reaches the main model. OneAdvanced went from prototype to production, deploying over 50 agents in three weeks, and achieved ISO 42001 certification for AI governance.
- Abbreviations
- AWS = Amazon Web Services — Amazon Web Services
- UK = United Kingdom — Соединенное Королевство
- LLM = Large Language Model — большая языковая модель
- RAG = Retrieval Augmented Generation — генерация с дополнением поиском
- ECS = Elastic Container Service — сервис эластичных контейнеров
- S3 = Simple Storage Service — сервис простого хранения
- SaaS = Software as a Service — программное обеспечение как услуга
- ISO = International Organization for Standardization — Международная организация по стандартизации
Source: AWS ML blog —
original