Built from teacher need, not AI hype.
We didn't trust the questions AI tools produce, so we built SayaLab to do the pedagogical work most tools skip. It reads your material first, then validates every question against your objectives and the right Bloom's level, and shows you the evidence. A teacher with decades in the classroom keeps the pedagogy honest.
It started with a builder's problem, not a teacher's.
We build software, and we kept seeing the same thing from AI tools: the questions they generated looked fine and often weren't. They clustered at recall, the wrong answers were easy to spot, and nothing mapped back to what the material was meant to teach. We didn't trust the output, and we doubted any educator should either.
So we built the part the tools skip. SayaLab reads your material before it writes anything, mapping objectives and checking the content can actually support them. Then it independently validates every question: the level it tests, its match to your objective, and the plausibility of its distractors. It shows you that evidence on each one.
The conviction underneath it is simple. No educator should have to ship an assessment they cannot stand behind. AI is the means, not the pitch: the model is a commodity, and the pedagogy on top is the product.
Saya means teacher.
The name comes from saya (ဆရာ), Burmese for teacher: a term of respect for the person at the front of the room. That is who we built this for, and who keeps us honest.
Builders who've done this before, kept honest by a teacher.
This isn't our first edtech. We built and scaled a learning platform past 100,000 educators and learners. We are doing it again, for assessment, and because we are builders and not teachers, we don't guess at the pedagogy.
A veteran teacher keeps us honest.
We are a product owner and an AI engineer, not teachers, so we didn't guess at the pedagogy. A teacher with more than 30 years in the classroom reviews the method and the questions it produces, and keeps us honest about what “good” means. The pedagogy is checked by someone who has taught, not assumed by people who haven't.
The principles we build to.
- Evidence over assertion — Trust is earned by inspection, not by a claim. Every question carries its own evidence: the Bloom's level we verified, the objective it maps to, how plausible its distractors are. You can see why each one holds up.
- You stay the author — A workshop, not a black box. You see why any question is flagged, and you can regenerate or edit any single one yourself. Nothing is locked in, and the assessment is yours to stand behind.
- Pedagogy first, AI second — We don't sell AI. We sell assessments educators can stand behind, shown with the evidence and the pedagogy behind them. The model is the means; the pedagogical work is the product.
- Honest about the limits — We name the research, use real numbers, and say what the evidence does and does not show. When your material is too thin to test something well, we tell you, instead of inventing a question anyway.