How our AI works
Last updated: 11 October 2026
Vidya uses AI to explain ideas, write practice questions, plan your study and estimate your readiness. AI can be wrong, so we label AI results, show why you're seeing them, and keep a record so we can check quality over time.
Your material stays yours
Anything you upload, and anything generated from it, is only visible to you. It is never shown to other learners.
Giving feedback
Tap Was this useful? under an AI answer, or Report a problem with this question under a question. A question reported by two people is hidden while we review it. You can also enter your real exam result in My Exams so we can check how accurate our score estimates are. Questions: support@atriverse.com.
The models
Admissions & interview coach
Feedback on application documents and interview answers.
Good for: Improving the learner's own drafts.
Not for: Writing applications on the learner's behalf.
Known limits: Generic advice for niche programmes.
Aspirations & career guide
Suggests paths from a learner's goals.
Good for: Exploration.
Not for: Career decisions without a counsellor.
Known limits: Salary and cutoff data may be dated.
Blind answer verification
Re-solves generated questions without the key to catch wrong answers.
Good for: Quality gate before generated questions are offered for review.
Not for: Replacing human review.
Known limits: Agrees with wrong keys when the source is ambiguous.
Design studio critique
Critiques uploaded art/design work for NIFT/Pearl/fine-arts prep.
Good for: Practice feedback on the learner's own work.
Not for: Grading for admission.
Known limits: Taste is subjective.
Exam readiness & score prediction
Predicts a p10/p50/p90 score range and pass probability for a private exam.
Good for: Guide what to study next and show a realistic score range.
Not for: Admissions, hiring or any decision about a person other than the learner. Not a guarantee of an exam result.
Known limits: Few responses give wide ranges; assumes item difficulty estimates are accurate; cannot see test-day conditions.
GK feed summariser
Summarises current-affairs items for GK practice.
Good for: Daily GK reading.
Not for: News of record.
Known limits: Depends on sources found.
A person reviews this output before it is used.
GMAT phased tutor
Diagnose, teach, check, stretch and reflect on one GMAT micro-skill.
Good for: GMAT practice with original questions.
Not for: Reproducing official GMAT items.
Known limits: Grading of short answers can be lenient.
Lesson writer & judge
Writes and accuracy-judges lessons and concept content.
Good for: Drafting lessons that are reviewed before going live.
Not for: Unreviewed publication.
Known limits: Judges share the writer's blind spots.
A person reviews this output before it is used.
Life rituals insight writer
Writes reflections from ritual completions.
Good for: Encouraging reflection.
Not for: Mental-health assessment.
Known limits: Generic for sparse data.
Life-skills role-play coach
Runs practice conversations and scores them.
Good for: Practising interviews and conversations.
Not for: Therapy or crisis support.
Known limits: Scores are subjective.
Material embeddings
Embeds a learner's source chunks for retrieval.
Good for: Finding passages in the learner's own material.
Not for: Cross-user search.
Known limits: Retrieval misses paraphrases.
Material extraction & blueprint
Transcribes uploaded pages and proposes modules/objectives.
Good for: Turning a learner's files into searchable text and a draft blueprint.
Not for: Anything beyond the uploader's own workspace.
Known limits: Handwriting and complex tables can be misread.
A person reviews this output before it is used.
Mock exam scoring
Scores timed mocks and updates per-module ability.
Good for: Scoring the learner's own practice mocks.
Not for: Official certification of a score.
Known limits: Ability update assumes a 1-parameter-style model.
Past-paper method cards
Writes method cards and solutions for past papers.
Good for: Explaining how to solve published questions.
Not for: Claiming official solutions.
Known limits: Can mis-solve hard items.
A person reviews this output before it is used.
Question generation from uploaded material
Writes draft practice questions, cards and comparisons from a learner's own uploaded material.
Good for: Private practice for the learner who uploaded the material, after review.
Not for: Sharing with other learners; official exam reproduction.
Known limits: May misread scanned pages; can produce wrong answer keys; only as good as the source.
A person reviews this output before it is used.
Shared practice question generator
Generates and judges original shared practice questions for public lanes.
Good for: Shared practice banks after automated judging.
Not for: Reproducing copyrighted official items.
Known limits: Difficulty estimates are rough until calibrated.
Study-plan builder
Schedules daily study by weight × weakness up to exam day.
Good for: Suggesting a daily plan the learner can edit.
Not for: Guaranteeing coverage of an official syllabus.
Known limits: Assumes stated minutes per day; protects hand-written days.
Teach-back grader
Scores a learner's explanation of a concept.
Good for: Formative feedback.
Not for: Grades that affect anything outside the app.
Known limits: Rewards fluent wording.
Tutor answers & explanations
Answers learner questions and explains concepts Socratically.
Good for: Learning support for students, including minors, under the safety prompt.
Not for: Medical, legal, mental-health or financial advice; role-play; homework submission.
Known limits: Can be confidently wrong; web mode depends on search quality.
Video & trailer script writer
Storyboards and scripts for explainer videos and trailers; image prompts.
Good for: Producing lesson media that is reviewed in Video Studio.
Not for: Unreviewed publication.
Known limits: Visual accuracy varies.
A person reviews this output before it is used.