Model answers at three depths
The same question answered for 1, 5 and 10 marks, written to the mark scheme your examiner actually uses, so you write the right amount.
Applied AI for education
Rote it, learn it, or scroll it. You pick how you study, and the exam is always the target.
Visit notesharu.com
Nepali students revise from photocopied notes nobody checked and past papers nobody solved properly. Some have an exam next week and need to pass. Others want to genuinely understand the subject. Almost every study app picks one of those students and ignores the other. notesharu serves both from the same material, and lets you choose which one you are today.
What you get
The same question answered for 1, 5 and 10 marks, written to the mark scheme your examiner actually uses, so you write the right amount.
Years of question papers, worked through properly rather than left as a scanned PDF you cannot search.
Every topic tied to the real TU or NEB syllabus, so you can see exactly what is examinable and what is not.
Cards, cloze deletions and quizzes generated from the answers themselves, so revision follows the material.
An FSRS scheduler decides what to show you and when, so things stick instead of being forgotten by Friday.
Concepts taught step by step and visually, for when you want to understand rather than memorise.
30-second narrated explainers, for revising the way you already scroll.
Photograph any page of your notebook and get it turned into searchable, structured study material.
Every quiz earns XP. It sounds small, but it is the difference between opening the app and not.
A steady stream of the next useful thing, so you never sit looking at a blank search box.
TU BSc CSIT, BIT, and NEB Class 11 and Class 12.
Ask any question and get an exam-ready answer drawn from real solved papers.
Behind the features
Every feature above is powered by something we built. Here is what is actually running.
The assistant
Ask a question in plain language and get an answer assembled from digitised past papers, real teacher and student notes, and the official syllabus, rather than from a model's general memory of the subject.
Grounding matters more here than anywhere else we work. A confidently wrong answer does not just annoy someone, it costs a student marks.
Document AI
The material starts as photographs and scans of handwritten and printed notes, much of it in Devanagari, which general OCR handles badly.
A layout-aware pipeline turns those pages into structured, searchable content, scores its own confidence, and sends anything doubtful to a human. The same pipeline powers Scan, so students can digitise their own notebooks.
Written for the mark scheme
A 1-mark answer and a 10-mark answer are different pieces of writing, and students lose marks writing the wrong one. We generate all three depths from the same source material, matched to how papers are actually marked.
Every generated answer is checked back against the source paper before it goes live.
Made to stick
Reading something once does almost nothing. Cards, cloze deletions and quizzes are derived automatically from each answer, then scheduled with FSRS so you see them again exactly when you are about to forget.
Cram mode reorders everything by exam likelihood and your weakest cards when the exam is days away.
Ask AI
General chatbots will answer any exam question confidently and sometimes wrongly, with no idea what your university asks or how it marks. notesharu answers from thousands of solved past papers and the official syllabus, tells you which paper an answer came from, and writes to the mark scheme. That is the whole difference.
Every answer follows the official TU and NEB syllabus and is checked against the real papers. Launched July 2026.