Gyanis is an AI-powered study OS for K-12 students. It offers Flow for AI study planning and adaptive practice, Aura for student mental health support with licensed psychologists, and Vibe for Ask-an-Expert tutoring. Gyanis helps students prepare for CBSE, IB, IGCSE & MOE Curriculum with personalised study plans, drills tagged to their syllabus, and a wallet-based credit system that runs across all four pillars.

Methodology

The system picks the next question, the next mindmap, and the next Gameplan label using rules we can name. This page lists those rules. Researchers, teachers, and regulators should be able to read it and challenge it.

Data sources

Official syllabus documents for each board we serve, including: CBSE, IB Diploma, IGCSE, UAE MOE.
Past paper archives, licensed or publicly available through the relevant board.
Teacher-authored content, commissioned for topic weightings, common mistakes, and marking scheme interpretation.
Aggregate, non-identifying signals: answered questions, time on task, and topic mastery. Identifying fields are removed before any signal enters the calibration layer.

How the system picks what comes next

Difficulty bands. Every question carries an ELO-style rating. The student carries one too, per sub-topic. Two correct answers in a row move the student up a band. One incorrect answer at the new band moves them back down.
Next best action. A lightweight rule engine decides whether to drill a weak area, revise a recent topic, or push into new material. The choice is bounded by the daily time budget the student set in Gameplan.
Confidence smoothing. Scores are smoothed nightly over a seven-day window, so one rough session does not move a topic from Conquered back to Mission.
Human override. Any teacher rolling out Gyanis can pin a topic, lock a mock, or reset the personalisation signals for a class from the admin panel.

How fast things happen

Mindmaps generate in under ten seconds. A typical Year 11 chapter produces a tree of twelve to thirty nodes, each with a three-line summary.
Drills are tagged on creation. Every question carries five tags: board, class, subject, unit, and topic. Filters work at the topic level, not the chapter level.
Gameplan labels update daily. Mission, Battle, Conquered, and Replay move based on the last three days of drill accuracy and the five days of drill speed.
Pulse refreshes at minute granularity. Time on task, accuracy by topic, and predicted score against the target are the same data points a student, a parent, and a teacher see.

How we check the system is working

Twice a year, we publish a benchmark report: the system's predicted difficulty versus actual student performance on held-out past paper items from CBSE, IB Diploma, IGCSE, UAE MOE. Aggregate numbers go on the Gyanis blog. Schools on a data-sharing agreement receive the cohort-level breakdown.

Aura is not algorithmic

Aura is a directory of DHA/DOH licensed coaches. Sessions are conducted by people, not models. The platform does not diagnose, does not prescribe, and does not substitute for clinical judgment. The only automation in Aura is the booking flow and the mood prompt, both labelled in the product.

What we will not do with student data

Student data is not used to train third-party models without explicit opt-in.
No interest-based ads are served to students. No student data is shared with advertisers.
UAE PDPL (Personal Data Protection Law) compliant. Parents can request a copy or deletion of a child's data via privacy@gyanis.ai.

Data sources

Official syllabus documents for each board we serve, including: CBSE, IB Diploma, IGCSE, UAE MOE.

Past paper archives, licensed or publicly available through the relevant board.

Teacher-authored content, commissioned for topic weightings, common mistakes, and marking scheme interpretation.

Aggregate, non-identifying signals: answered questions, time on task, and topic mastery. Identifying fields are removed before any signal enters the calibration layer.

How the system picks what comes next

Difficulty bands. Every question carries an ELO-style rating. The student carries one too, per sub-topic. Two correct answers in a row move the student up a band. One incorrect answer at the new band moves them back down.

Next best action. A lightweight rule engine decides whether to drill a weak area, revise a recent topic, or push into new material. The choice is bounded by the daily time budget the student set in Gameplan.

Confidence smoothing. Scores are smoothed nightly over a seven-day window, so one rough session does not move a topic from Conquered back to Mission.

Human override. Any teacher rolling out Gyanis can pin a topic, lock a mock, or reset the personalisation signals for a class from the admin panel.

How fast things happen

Mindmaps generate in under ten seconds. A typical Year 11 chapter produces a tree of twelve to thirty nodes, each with a three-line summary.

Drills are tagged on creation. Every question carries five tags: board, class, subject, unit, and topic. Filters work at the topic level, not the chapter level.

Gameplan labels update daily. Mission, Battle, Conquered, and Replay move based on the last three days of drill accuracy and the five days of drill speed.

Pulse refreshes at minute granularity. Time on task, accuracy by topic, and predicted score against the target are the same data points a student, a parent, and a teacher see.