Juan Herreros Elorza

Juan Herreros Elorza is a platform engineering leader building infrastructure that serves software developers and AI agents in regulated financial environments. At Banking Circle, he helped develop Atlas, an internal platform combining Kubernetes-based compute, infrastructure provisioning, messaging, and observability.

By 2023, Herreros Elorza was Banking Circle’s platform engineering tech lead, advancing self-service infrastructure as code through reusable Terraform modules, standardized Azure DevOps pipelines, and clearer platform governance. His work on GitOps and automated compliance extended those principles to controlled releases and feature flags in regulated banking. He subsequently explored Crossplane and Helm infrastructure patterns and developed Mini-Atlas, an experimental internal developer platform featured on his personal website.

Building platforms that agents can actually use

• Self-service without human handoffs: Agents cannot navigate undocumented deployment procedures, request infrastructure from specialists, or wait for informal approvals. Platforms must expose complete, independently executable workflows.

• API-first infrastructure with enforceable controls: Well-defined APIs, command-line tools, and MCP servers give agents discoverable capabilities, structured inputs, authentication, authorization, and predictable provisioning paths.

• Local-first validation and machine-accessible observability: Rapid local checks prevent wasted pipeline cycles; queryable logs, metrics, and traces let agents verify that deployments actually worked.

• Documentation and guarded contribution: Repository instructions, reusable skills, and discoverable platform documentation help agents follow established build, testing, and deployment conventions. Enforced security and compliance policies keep platform ownership accountable as AI-assisted contributions expand.

In his AI Engineer Europe session, Herreros Elorza grounds agent adoption in measurable outcomes: delivery performance, reliability, support demand, and developer experience. His central practical insight is that autonomous agents expose organizational friction humans have learned to work around—and make fixing it unavoidable.