PrepAI (Interview Preparation App) — a project by Ahsan Mahmood
Interview-prep platform pairing ~900 hand-authored, browser-runnable coding drills across six language tracks with mock interviews, peer reviews, and a…
A question dump is cheap to build. A table of 900 questions is an afternoon of work, and it teaches nobody anything.
PrepAI is the expensive version of that idea.
About 900 drills are hand-authored across six tracks: JavaScript, TypeScript, HTML, CSS, Node.js and Node microservices, and hand-authored is the word doing the work in that sentence. Each one carries a starter, a hidden test suite and three tiers of progressive hints.
The hints are the part that took the time.
A hint that gives the answer has removed the exercise, and a hint that says try again is not a hint at all. Three tiers exist because the useful help sits somewhere between those two, and exactly where depends on the drill.
Then you press run.
The code executes in a sandboxed Web Worker inside your own browser, through CodeMirror or Monaco, and the result comes back immediately. There is no execution backend behind any of it.
Two editors ship rather than one. CodeMirror is the light one and Monaco is the editor most people already have muscle memory for, and asking a candidate to learn an unfamiliar editor while being timed is a strange thing to do to somebody.
That constraint is visible in the syllabus.
Every one of the six tracks runs in a browser. A drill in a language the browser cannot execute would need a server to execute it on, and there is no server here.
The language list and the budget are the same decision.
Around the drills sit the things that make preparation resemble the real thing. Mock interviews run three ways, with an AI interviewer, a peer or a mentor. Candidates exchange code reviews with each other, climb leaderboards and talk to a Gemini mentor running on Google's free tier.
Peer review is the part candidates avoid and the part that earns its place. Reading somebody else's solution to a drill you have just failed is a different kind of correction from a hint, and it costs the platform nothing to arrange.
Five roles share one product.
Candidates, mentors, instructors, HR reviewers and admins each get a workspace scoped to what they are there to do, which is a considerably harder exercise than putting five menus behind a single login and hoping people ignore the four that are not theirs.
Nothing underneath costs money to keep running.
Firebase's free tier holds the data. The Gemini mentor runs on Google's free tier and uploads go through FilesHub, with no paid execution backend and no Stripe integration anywhere in the product.
The plan set is Free, Pro at $8 and Ultimate at $14.
Those are the only figures on this page.
What it cost is the syllabus.
Six tracks that all run in a browser is a real boundary, and somebody preparing for a Python or a Java interview is not served here. The alternative was a server willing to execute arbitrary code sent by strangers, which is a different product with a different bill and a much larger security surface.
That trade was made deliberately.
The second cost is authorship. About 900 drills, each with a hidden test suite and three tiers of hints, is not something anybody scrapes.
It is in progress rather than finished.
It is at interviewpai.aoneahsan.com, with a companion Chrome extension beside it.
A hidden test suite is only ever as good as the person who wrote it, and there was one of those.
What it does, and what that costs to build
- ~900 browser-runnable coding drills (6 tracks)
- Sandboxed JS Web Worker execution (zero cost)
- Dual editor (CodeMirror + Monaco)
- AI/peer/mentor-led mock interviews
- Peer reviews, gamification & leaderboards
- Gemini AI mentor (free tier)
- Five roles in one product
Built with
- React 19
- TypeScript
- Vite 8
- TanStack Router
- Radix Themes
- Firebase
- CapacitorJS
- CodeMirror
- Monaco
Worth knowing
- education
- interview-prep
- coding
- career
- ai
- react
Who built PrepAI?
Ahsan Mahmood built it working alone, and it is at interviewpai.aoneahsan.com. The drills, the hidden test suites, the sandboxed execution, the five role-scoped workspaces, the Firestore rules and the companion Chrome extension are one person's work. It is React and TypeScript on Vite with the React Compiler, TanStack Router for navigation, Radix Themes on the interface, CodeMirror and Monaco as the two editors and Firebase underneath, packaged with Capacitor for mobile.
What does PrepAI cost?
There are three tiers: Free, Pro at $8 and Ultimate at $14. Those are the figures the record carries and this page adds nothing to them. The reason a free tier can be generous here is that the expensive part of an interview-preparation product usually is code execution, and that runs in the candidate's own browser rather than on hardware anybody is renting. Payment does not go through Stripe, and there is no paid execution backend anywhere in the design.
What happens when I run a drill?
It runs in your own browser, inside a sandboxed Web Worker, and returns a result immediately. There is no execution backend, no queue and no waiting for a container somewhere to start. The editor is CodeMirror or Monaco, so the drill is written and run in the same place. This is also why the tracks are what they are: every one of the six runs in a browser, and a drill in a language a browser cannot execute would need a server to execute it on.
Why does PrepAI have five different roles?
Because interview preparation involves five different jobs, and each of them needs a workspace scoped to it. Candidates work through drills and sit mock interviews. Mentors and instructors run sessions and review work. HR reviewers look at candidates rather than at code, and admins look after the platform. Putting five menus behind one login would have been easier and would have left every role reading past four sets of controls that were not theirs, which is the version most products ship.
Does PrepAI work on iPhone?
No, there is no iOS build. PrepAI is a web application, so an iPhone reaches it in a browser exactly as any laptop does, and since the drills execute inside the browser itself very little is lost by doing that. The packaged mobile build is Android through Capacitor, and the companion extension is Chrome. There is no Apple Developer account behind this work and nothing here has been submitted to the App Store.
https://aoneahsan.com/projects/com.aoneahsan.interviewpai