Third-party cyber evaluations involving OpenAI models
openai.com·14h ago
Cloud & Infrastructure
Google Cloud Blog·Jun 15·5 min read
Architecting a trusted agentic platform with graph technologies: A Yahoo case study
TL;DR
Enterprises faced inefficiencies in digital media buying due to lengthy manual processes. Yahoo built the Seller Agent platform using Google Data Cloud's graph technologies to automate and govern media campaigns.
✦ Why It Matters
Engineers can leverage graph technologies to build efficient, accountable systems in various domains.
Key Takeaways
- Implement a knowledge graph to ground decisions in business reality.
- Use a context graph for creating an auditable memory of decisions.
- Adopt open protocols for transparency in agentic behavior.
- Leverage BigQuery Agent Analytics for real-time decision tracking.
How It Works
The Seller Agent platform operates through a multi-agent system where requests are processed by a planning supervisor agent. This agent decomposes tasks into specialized functions, utilizing a knowledge graph for real-time data and a context graph for tracking decisions.
The dual-graph architecture allows for rapid execution while ensuring that every decision is traceable and compliant with governance policies.