Build a multi-agent orchestration layer to unify disconnected AI systems.
Most companies eventually realize their isolated, single-purpose AI agents create a fragmented, frustrating user experience. An orchestration layer solves this by acting as a central brain that routes queries, shares context, and synthesizes answers into one coherent response. Building this project proves you can architect complex, production-grade AI systems, moving beyond simple API wrappers to tackle routing, latency, and multi-agent coordination.
The Brief
An online education platform runs several independent learner-facing AI agents (e.g., an admissions agent and a career advisor) built on different stacks. When a learner asks a complex question spanning both domains, the system fails to share context and delivers disconnected answers. Your job is to design and build an orchestration layer that sits above these agents. It must look up the learner's status from a CRM stub, route the query to the correct agents, and synthesize the outputs into a single, coherent response.
This project is as much about communication as it is about code. Before building the runnable prototype, you must write a strategic 2-page architecture memo defending your technical choices, outlining success metrics, and detailing how you would scale the system for 10,000 learners. The final deliverable includes both the code and the memo, demonstrating your ability to communicate architectural tradeoffs to leadership.
The Idea Behind It
This education platform is suffering from "agent sprawl." They need a builder who can step back from individual prompts and design a higher-level routing architecture that unifies their disparate AI systems and integrates with their CRM.
This project prepares you for senior technical roles because it tests system design over raw coding. It proves you can evaluate frameworks, manage handoff failure modes, and write a strategic memo defending your architecture to a CTO.
What You Will Build
- A 2-page strategy memo detailing your architecture shape, framework choices, success metrics, and top production risks.
- A functional orchestration prototype that routes a complex user query to multiple simulated downstream agents.
- A CRM integration stub that looks up user context before routing.
- A synthesis step that merges outputs from multiple agents into a single, coherent response.
- A technical README explaining how to run the system, what is stubbed, and the required steps to reach production scale.
Tests system design, strategic communication, and complex routing patterns far beyond basic LLM integrations.
Learn how to architect and orchestrate complex AI systems in the next AI Native Builder cohort.
Apply to Cohort