Poko
The problem.
Producing high-quality assessment content is notoriously slow and expensive. Methodologists draft materials, subject matter experts formulate questions, QA verifies accuracy—and none of this scales. Consequently, educational institutions struggle with duplicated content, inconsistent question quality, lack of repeatable processes, and difficulty in maintaining multiple difficulty levels.
What we built.
We engineered a modular platform where AI handles question generation, while humans retain control and validation roles. The content follows a strict hierarchy—course, lesson, card, and question—which effectively separates the core knowledge from its assessment, making all learning materials highly reusable.
- The user pastes a text or highlights a fragment, then selects the question type, quantity, and difficulty level
- The system injects variables into a prompt template and enforces a strictly structured JSON response from the AI
- Supported question types include true/false, multiple choice, and fill-in-the-blank
- Prompts are customizable but within safe constraints, allowing experts to fine-tune the system without breaking the logic
- Three versatile card creation modes: from highlighted text, from pasted text, or fully manual
Check out the ui.
The hard part.
This is a heavily backend-driven product where the UI controls complex logic rather than just serving as a storefront. It features role-based access, forms with conditional logic, saved prompt templates, parameterized LLM requests, strict output formatting, multiple content creation flows, document versioning via editing and archiving, and seamless exports to external systems. The primary user of this interface is not the end student, but rather the operational staff managing the educational process.
What changed.
- Content creation time was dramatically reduced without increasing headcount
- Output quality is rigorously controlled via standardized prompts rather than relying on individual authoring skills
- Unlocked multiple new monetization avenues: per-seat billing, pay-per-generated-question, enterprise licensing, and white-label options