From a school request to a product foundation

Kayan Assessment is an AI-powered platform requested by Kayan International High School to address the time and effort teachers spend creating, correcting, and reviewing curriculum-aligned assessments in Lebanon.

I am developing the product with the school while leading its product definition, architecture, and implementation. The work turns a high-friction institutional process into a structured SaaS workflow without treating AI output as a substitute for teacher judgment.

Designed for the domain

The platform is organized around the real standards, terminology, curriculum material, evaluation logic, and approval needs of its users. Structured domain content is ingested securely; AI assists with assessment work; and teachers retain review and approval authority.

That distinction shapes the product. It is not a general chat interface placed over uploaded files. It is a system of repeatable workflows, explicit roles, traceable activity, and quality feedback.

Building a compounding system

Each approved interaction can contribute to a growing structured dataset. Over time, that feedback can strengthen assessment quality, surface recurring curriculum patterns, and create an advantage that comes from real institutional usage rather than dependence on a single model.

The AI layer is provider-independent, supporting three interchangeable providers through one abstraction. Requests can be routed according to capability, cost, latency, and availability while the surrounding product logic remains stable.

Product foundation

The private MVP includes or is designed around:

• Multi-tenant organizations, users, roles, and permissions
• Structured curriculum, document, and file ingestion
• Secure object storage
• AI-assisted assessment creation, correction, and review
• Human approval workflows
• Administration, usage, and quality dashboards
• Structured feedback collection and audit-friendly history
• Cloud deployment and pilot onboarding

The architecture spans 12 interconnected modules, four role-specific experiences, 30+ structured entities and relationships, and 25+ end-to-end workflows. Approximately 80% of the private MVP architecture is implemented.

Lean enough to pilot, ready to grow

The initial infrastructure is designed to operate for under $50 per month during early pilots. Internal workflow runs complete successfully 99.5% of the time, with a 1.6-second median response time for non-AI APIs and a projected 65% reduction in the time spent on the core manual process.

The venture is preparing for pilot validation, prospective partnerships, and early-stage funding. Selected details of the workflow, data strategy, orchestration rules, and product logic remain private while the school collaboration and MVP develop.

My ownership

I translated the initial opportunity into a product, researched the workflow, designed the data and multi-tenant architecture, built authentication and authorization, created the AI orchestration and file-processing foundations, developed the user-facing application, and planned the feedback system and cloud path.

The project demonstrates product ownership at venture scale: finding a valuable problem, turning ambiguity into a system, building under strict cost constraints, and balancing rapid delivery with a defensible long-term data strategy.