Question banks don't know you
A thousand questions in a PDF are ordered by chapter, not by what you personally keep getting wrong. You spend your best hours re-solving things you already own.
PrepCelerator is an adaptive practice app for students preparing for JEE Main, JEE Advanced and NEET. It tracks how well you know every single topic in your syllabus, then picks your next question from that — not a random bank, not a chatbot guessing, but a scoring engine aimed squarely at whatever is costing you marks.
Why this exists
Most students aren't short of material. They're short of a reliable answer to where the next hour is best spent.
A thousand questions in a PDF are ordered by chapter, not by what you personally keep getting wrong. You spend your best hours re-solving things you already own.
A test tells you 62%. It doesn't tell you that Rotational Motion and Mole Concept accounted for most of what you dropped, or which one to fix first.
The question you got wrong three weeks ago is the one most likely to appear again — and the one you're least likely to remember to revisit.
PrepCelerator answers that question for you, every single time you open it.
How it works
Setup takes a few minutes once. After that the app decides what comes next, every time.
Tick off the chapters and topics you've actually covered. Anything marked not-learned is excluded from question selection until you're ready for it.
Infinite adaptive practice, a scoped chapter drill, or a full timed paper. Set a per-question timer if you want the pressure.
A drawing canvas sits beside every question, resizable and optionally carried over to the next one. Formulas render properly, not as raw LaTeX.
The server grades your answer, updates your mastery for that topic, schedules a revision if you missed it, and that new state decides what you see next.
The engine
It's a deterministic scoring function over your own history. That matters: it can be reasoned about, tuned, and trusted to behave the same way twice.
Question selection is a scoring function over stored mastery and attempt history. No LLM is anywhere in that path — the same state always produces the same reasoning, and it can be explained rather than guessed at.
Every topic carries its own IRT-style mastery score, updated after each answer. Being strong at Kinematics tells the engine nothing about your Vectors.
Candidate questions are sampled with weight, not ranked and repeated — so weak topics come up far more often without the session turning into the same five questions.
A question you just saw is held back for a set number of attempts and hours, so a weak topic gets new questions rather than the same one on a loop.
The scoring weights live in a database table. Tuning how aggressively the engine chases weakness is a config change, not a redeploy.
Mastery is path-dependent, so removing a session replays your surviving attempts through the identical formula rather than faking a subtraction. Your numbers stay true.
Practice modes
Scoped drills, open-ended practice, or a full paper under exam conditions — the same mastery model powers all of them.
The default. Questions picked for you across your entire syllabus, one at a time, for as long as you want to keep going. Every answer re-aims the next question.
Unlimited · whole syllabus · adaptive
A focused 20-question paper for one subject, under real exam timing and marking.
20 questions · 1 hour · marked
Pick specific chapters and drill only those. Still adaptive inside the scope you chose.
Your chapters · adaptive
Narrower than a chapter. Select individual topics when you know exactly what's shaky.
Your topics · adaptive
Extra practice restricted to the topics your mastery data says you're struggling with.
Auto-selected · adaptive
Revisit questions you got wrong, resurfaced on a spaced schedule so they stick.
Spaced repetition
A complete 75-question paper — 25 each of Physics, Chemistry and Maths — under exam conditions.
75 questions · 3 hours · +4/−1
Exam realism
Which means no hints while the clock runs, marking that matches the real rules, and a report afterwards that tells you something you can act on.
| Paper | Length | Duration | Marking |
|---|---|---|---|
| Full Mock Test | 75 questions | 3 hours | +4 / −1 |
| Subject-wise Test | 20 questions | 1 hour | +4 / −1 |
During a mock or subject test you don't find out whether an answer was right — same as the real thing. Previous-question review is disabled too.
MCQs carry +4/−1 and numerical questions +4/0, resolved per question type rather than one flat scheme for the whole paper.
If the clock runs out with nothing selected, it's recorded as skipped and scored zero — not as a wrong answer costing you a mark.
Every question of a mock test is chosen up front, so the paper can't shift under you mid-test.
Subject-wise marks and accuracy on screen, plus a downloadable report with per-topic strengths and weaknesses for that session alone.
A topic is only called a strength above 60% accuracy and a weakness below 40%. If there isn't enough signal, the report says so instead of inventing a verdict.
AI, in its place
That line is drawn deliberately. A language model is genuinely good at rephrasing a hard idea; it has no business choosing your syllabus or marking your paper.
Ask for a simpler explanation or a full step-by-step walkthrough of any question you've already attempted.
Pick a chapter and get it explained from scratch, with maths rendered properly — headings, steps and formulas, not a wall of text.
Set your explanation language once in your profile and every AI explanation follows it.
Once a week, a summary of what you practised, which subjects are genuinely exam-ready and which aren't — calibrated to JEE cutoffs, not graded on a curve against yourself.
Thumbs up or down on any AI response. Low-rated answers land in a review queue instead of quietly staying wrong.
Explanations unlock only after you’ve attempted the question — you can’t read your way to the answer before trying it.
Concept animations
Short clips that explain one concept properly, in a vertical feed that works exactly the way your thumb already expects. The difference is what decides the order.
Type what you’re stuck on in plain words. Full-text search across every clip’s title, topic and tags.
With no search query, the feed sorts against your own mastery data — weakest first. An exam-prep feed should surface what you need, not what got the most likes.
Only the clip you’re watching loads. A full pass through the feed costs less than a single photo on most apps.
Progress
Not a wall of dashboards. A short answer to how you're trending, which topics are dragging, and what to open next.
Click any row in the app to start practising that topic immediately.
Charts shown are illustrative.
Session-by-session accuracy over time, so you can see whether the line is actually moving.
Roll your mastery up by subject, chapter or topic — weakest first — and click any weak row to start practising it immediately.
Every mock test you've taken, plotted, once you've taken your first.
A daily practice streak and a milestone ladder from your first five sessions up to five hundred, both derived from real activity — nothing is awarded for showing up.
Plan your study day by day. Add what you want to get done, with or without a time, and tick it off as you go. Anything still due today waits for you on the home page.
A vertical feed of short clips explaining concepts — searchable, and ordered by your weakest topics rather than by what's popular.
Built to be trusted
Both matter more than they sound. One protects your practice history; the other is what makes your accuracy figures mean anything at all.
Solutions and correct answers are unreadable by a student account at the database layer. They're released only after you've attempted the question — not hidden in the page waiting to be found.
Your answer is graded, your mastery updated and your revision scheduled inside one server-side transaction. A client can't report its own result.
Every table enforces per-user access at the database, not in application code. Your attempts, mastery, tasks and reactions are readable only by you.
The app holds no privileged service key anywhere. Every server-side request runs as you, with your permissions, and nothing more.
You can create your PrepCelerator account with an email and password, or with Google sign-in. If you choose Google, we request three things: your name, your email address and your profile photo. The name and photo are used to personalise your account; the email address identifies it and is how password resets and account notices reach you.
That is the entire scope. We never request access to your Gmail, Drive, Contacts or Calendar, we never post anything to your Google account, and we do not sell your data or run advertising on the platform. Your practice history stays readable only by your own account.
Pricing
Every feature, not a stripped-down sample: the adaptive engine, all seven modes, mock tests and progress tracking. No card required to start.
Every feature, for your first month. No card required.
₹0for your first month
For the stretch where every mark counts.
₹2,499per year
Works out to ₹208 a month.
To be straight with you: Pro isn’t live yet — there’s no billing in the app today and no launch date set, so nothing can be charged automatically and no payment method is held. You’ll be told well before that changes, and your practice history carries over when it does.
FAQ
Including the ones that aren't flattering.
Not yet — PrepCelerator is invite-only while the question bank grows and we fix what early students run into. Join the waitlist with your email and we'll let you in as places open up. There's no charge and no card at any stage of that.
You get every feature — all practice modes, the adaptive engine, mock tests, progress dashboards and PDF reports — free for your first month, with no card required at signup. After that a Pro subscription is ₹2,499 a year. Billing isn't live yet, so nothing can be charged automatically and you'll be told well before that changes.
It's built for JEE Main, JEE Advanced and NEET. The question bank currently follows the real NCERT Class 11 chapter and topic structure for Physics, Chemistry and Maths, plus JEE-specific additions like Mathematical Tools and Basic Mathematics. Class 12 chapters are being added.
Two sources, kept clearly apart: originally-authored questions written for this app, and genuine previous-year questions from published exam papers. Each question records which of the two it is, so you always know whether you're solving a real past paper question.
No. The part that decides what you practise is a deterministic scoring engine over your own mastery data, with no language model involved at all. AI is used only to explain things after the fact — explanations, a concept explainer and a weekly recap — and never to choose or grade a question.
Not unless you want it. A drawing canvas sits beside every question with pan, zoom and undo, resizable against the question pane, and you can choose to carry your working over to the next question.
Yes. The app is responsive down to small phone screens and installs to your home screen as a progressive web app. It needs a connection — there's no offline question bank.
Each topic holds an IRT-style mastery score updated on the server after every answer, weighing whether you were right, how long you took, and how difficult the question was. It's a running value rather than a simple percentage, which is why the app recomputes it honestly if you ever delete a session.
There's no date yet, and billing isn't wired up. Whenever it does launch, your existing practice history, mastery and tasks carry over untouched — nothing you've built up is lost, and you won't be charged without being told first.
PrepCelerator is invite-only for now, while the question bank grows and we fix what early students find. Leave your email and we’ll let you in as places open up. Your first month is free and there’s no card at any point.
Already have an account? Log in