Tech Lead |Software Architecture & Production Systems

A 30-month study of 26,811 students found that generative AI raised homework scores while exam scores fell. The gap isn't a fluke—it's evidence that homework score has stopped being a valid proxy for learning, with direct consequences for how EdTech products measure engagement and how governance frameworks define risk.

As I explore AI governance through formal coursework and implementation context, I'm identifying specific gaps where educational chatbots introduce operational and compliance challenges that standard enterprise AI discussions often overlook.

What I'm learning about governance in AI education: why decision architecture has to precede technical architecture.

A prompt injection attack on an educational AI system is not just a security incident. Under EU AI Act Article 9, it is a risk your management system should have anticipated, documented, and mitigated before deployment.

Naive RAG retrieval relies on vector similarity alone. This fails silently when documents are semantically similar but contextually irrelevant. Re-ranking adds a second filtering gate that evaluates actual relevance.

RAG isn’t a single pattern, but a family of architectures. And choosing the wrong architecture isn’t a performance problem, but a design flaw.

The gap between the architecture diagram and the working system , documented from the inside.