03 June 2026 —
10:15 to
11:00 on
Web Stage
Large-scale web applications often start as monoliths and gradually evolve into distributed architectures. On the backend, microservices are well established — but the frontend frequently remains a bottleneck where multiple teams deliver through the same codebase and tooling.
Microfrontend architectures address this by slicing applications along business context boundaries into self-contained full-stack components: a UI built as a Custom Element with Shadow DOM, a lightweight backend-for-frontend for data aggregation and transformation, and its own CI/CD and infrastructure-as-code — all owned by a single team. Making these components work together without creating tight coupling requires discipline. Drawing on event-driven architecture patterns — publish/subscribe, choreography, orchestration — we can build an integration layer where microfrontends communicate through CustomEvents and HTML attributes, with an orchestrator that owns the workflow and translates between message semantics. The OpenMFE standard formalises this through manifest files that describe each component's configuration attributes, events, and data structures as machine-readable interface contracts using JSON schema. This architecture was designed for human teams, but its properties — formal contracts, enforced encapsulation, self-contained modules with clear boundaries — turn out to align remarkably well with the practical constraints of AI coding agents: finite context windows, no implicit architectural knowledge, and no ability to infer boundaries that aren't explicitly defined. We will explore why this kind of modular, contract-driven architecture makes applications evolvable not just by human developers, but by AI agents — and why that evolvability is becoming the defining architectural challenge ahead
Read More...
By
Alexander Günsche
[Senior Solutions Architect — Amazon Web Services]