이미지: METAL LAB 생성
Summary
- F1 built a Data Accelerator with AWS based on Amazon Bedrock AgentCore
- Onboarding new data sources shrank from up to 8 weeks to about 40 minutes of code generation
- The system delivers end-to-end observability, including automatic schema-change detection and correction and data lineage tracking
- 적용 기업
- Formula 1(F1), AWS
- 핵심 기술
- Amazon Bedrock AgentCore 기반 에이전틱 AI
- 온보딩 시간 변화
- 최대 8주 → 약 40분 코드 생성(배포는 별도 소요)
- 이전 백로그
- 신규 소스 12개 통합에 18개월 대기
- 주요 기능
- 스키마 진화 자동 감지·수정, 통합 데이터 접근, 근본원인분석(RCA)
Formula One (F1) has partnered with AWS to build a "Data Accelerator" that automates its marketing technology (MarTech) data platform, "Customer 360," according to the AWS ML Blog. The system uses agentic AI running on the Amazon Bedrock AgentCore runtime to cut the time needed to onboard new data sources from up to eight weeks down to roughly 40 minutes of code generation work (deployment takes additional time separately).
A manual bottleneck
Chris Roberts, IT Director at F1, said that "every new data source required six to eight weeks of manual engineering, and integrating just 12 sources had built up an 18-month backlog." As data flowed in from ticketing partners, streaming services, sponsorship activation feeds, social media, and merchandise systems, schema mapping, pipeline construction, data quality checks, and GDPR classification were all done manually.
Solved through five workstreams
The Data Accelerator consists of five workstreams: agentic source onboarding built on Bedrock AgentCore, automatic detection and correction of schema evolution, unified data access via Amazon SageMaker Unified Studio, root cause analysis (RCA) tools and observability based on a context graph, and a capability to automatically identify failures that can be resolved through code fixes. Matt Kemp, F1's head of data operations, explained that the team needed a repeatable and robust solution, and Roberts said, "For the first time, we have end-to-end visibility that includes data lineage and root cause analysis—not just dashboard alerts."
