이미지: AWS ML Blog
Summary
- PDI Technologies built the 'PDI Brew' platform on Amazon Bedrock and AWS Lambda
- Non-developers describe a tool in plain language and an SSO-protected multi-tenant app is automatically generated
- A planning agent turns intent into a manifest, and a Lambda provisioning agent creates the AWS resources
- 개발사
- PDI Technologies
- 플랫폼명
- PDI Brew
- 기반 기술
- Amazon Bedrock, AWS Lambda
- 회사 규모
- 직원 약 4,000명, 200여 개국 20만여 고객처
- AI 기능
- 챗, 요약, 분류 등 거버넌스 적용된 Bedrock 기반 기능
PDI Technologies, a software company serving the convenience store and fuel distribution industry, introduced its self-built agentic internal app deployment platform 'PDI Brew' via the AWS blog. The company said that when a non-developer employee describes a desired tool in plain language, a fully provisioned multi-tenant web app is generated within seconds, with no need for Git usage, terminal operations, or DevOps knowledge.
Dual Planning-Provisioning Agent Architecture
PDI Brew operates on a two-stage agent architecture. A planning agent converts the user's intent into a structured manifest, using skill features of AI assistants such as Claude, ChatGPT, and Claude Code, or by invoking Amazon Bedrock models. A provisioning agent running on AWS Lambda then breaks down this manifest to classify the workload, select the necessary tools, and orchestrate the creation of underlying AWS resources in a single request.
Because every generated app inherits the same platform, they come with enterprise SSO, scoped IAM, HTTPS, and centralized observability by default. Without developers having to manage separate model endpoints or API keys, each app can optionally use governed AI features such as chat, summarization, and classification powered by Amazon Bedrock.
PDI Technologies said it has a 40-year history with approximately 4,000 employees, and serves around 200,000 customer locations in more than 200 countries. The company explained that it built the platform out of the recognition that existing deployment pipelines were blocking non-developer teams from launching small-scale tools.
