Mauro Luchetti is an enterprise AI architect and AI Center of Excellence leader whose work at Quantyca focuses on making agent systems secure, discoverable and governable across large organizations. His approach combines knowledge graphs, shared platform infrastructure and detailed dependency tracking to prevent distributed AI development from becoming an operational liability.
In 2019, Luchetti was a Quantyca data engineer writing about web-service authentication and authorization. He argued for separating authentication from application logic, both to prevent unauthorized requests from consuming service resources and to make security infrastructure independently adaptable.
By 2022, he was working as a data architect on Arcese’s event-driven logistics platform, supporting near-real-time shipment monitoring through a digital integration hub. In 2024, his responsibilities encompassed enterprise generative AI strategy, including proprietary data, retrieval-augmented generation and knowledge graphs. With Andrea Gioia, he subsequently explored how enterprise knowledge graphs and domain-specific data products could give multiple agents a shared, coherent understanding of organizational information.
Making enterprise agents governable
Working with Sonny Merla of Amplifon and Mattia Redaelli of Quantyca, Luchetti helped develop an architecture for Amplifon’s AmplifAI initiative built around a centralized gateway and interconnected registries. The AI Engineer Europe presentation described an evolving platform; its demonstrated catalog contained sample data and had not yet entered production.
His contributions center on four practical mechanisms:
- Governed AI gateway: A unified model endpoint combines Microsoft Entra ID authentication, configurable spending limits and centralized monitoring, making security and budget controls shared infrastructure.
- Private MCP server registry: Internal tools and approved public Model Context Protocol servers receive ownership, environment, authentication, cost and use-case metadata, enabling accountability and impact analysis.
- Automated agent discovery: Agent-to-Agent registries capture agent cards describing capabilities, endpoints and authentication requirements; deployment pipelines publish those records automatically so teams can discover and reuse existing services.
- Use-case lineage: Linking business applications to their underlying agents, tools and models makes shared dependencies visible and helps teams identify which systems a model change or service outage could affect.
For Luchetti, enterprise AI governance depends on giving developers reusable infrastructure while preserving clear ownership, auditable access and visibility into how intelligent systems interact.
