An integrity-first coding-assessment platform for hiring and campus drives. Candidates code inside your own editor while you watch live; afterwards you get a ranked board, evidence-backed integrity flags, and a defensible hire recommendation per candidate — remote take-home or on-site.
A coding test gives you a leaderboard. A leaderboard lies in both directions.
- Cheaters float to the top. Pasted editorial solutions, a second device, a friend on a call, an LLM in another tab — the score looks great, the candidate can't reproduce a line of it in the room.
- Real talent gets buried. The careful problem-solver who wrote clean, original code under pressure ranks below the copier and never gets the call.
So the score you trust most is often the one you should trust least — and the hire you'd actually want walks away. You need to know two things a raw board can't tell you: is the work their own, and who is genuinely good?
Aerele Proctor turns a coding round into defensible evidence and a recommendation you can stand behind in a review — not just a number.
- Candidates stay in your editor. Registration, identity confirmation, the full coding workspace, and Run/Submit all happen inside one app you host. No tab to a third-party site, no leaving your environment.
- You watch the whole room from one console. Live status, screen and camera streams, integrity signals, and remote actions — all in a single admin view, for one candidate or a thousand.
- It catches cheating with evidence, not theatre. Fullscreen enforcement, the recorded screen, tab/recording/IP signals, and a deterministic post-exam analysis that clusters copied code, spots editorial pastes, and flags the "solved a hard problem on the first try" pattern.
- It surfaces the talent a leaderboard hides. A calibrated engine scores genuine problem-solving on its own axis — so the honest, capable candidate rises whether or not they topped the raw board.
- It ends with a recommendation you can defend. Each candidate gets a hire recommendation backed by a FOR / AGAINST argument and the underlying evidence — something a reviewer can interrogate, not a black-box verdict.
Aerele Proctor is built on one principle: it gives you evidence to review, never a sentence to execute. Integrity and talent are scored on separate axes that are never averaged — a flagged score is never silently "corrected," and a clean candidate is never quietly demoted. Calibration is deliberately conservative: an honest candidate is ~never excluded, only conclusive copying excludes, and a single shared problem is a desk-check note, not a disqualification. Treat every flag as triage for a human — a supervised pen-and-paper round or a reviewer is the ground truth, and the platform is designed to make that round sharper, not to replace it.
It is also honest about what a browser can do. This is real, browser-based proctoring — evidence collection and human review, not lockdown spyware. It will not enumerate your other tabs, read your OS clipboard, or police a second monitor; no web app can. What it does do — hold fullscreen, record the entire screen, watch the live feed, track the signals that matter, and turn all of it into a reviewable case — is enough to run a coding round you can trust and defend.
- Engineering hiring teams running take-home or on-site coding rounds who need a shortlist they can defend — not a leaderboard they have to second-guess.
- Campus & placement drives assessing hundreds of students at once, where the goal is to find the genuine problem-solvers and reconcile against a pen-and-paper round.
- Anyone who wants to own their assessment stack — self-hosted on your own Google Cloud, your data, your editor, no per-seat SaaS bill.
Use it remote (candidates self-serve with an access code from anywhere) or on-site (a tokenized per-room invigilator portal for floor staff) — the same platform, the same evidence, either way.
The fastest way to understand the product is to run the UI. Demo mode runs the entire experience on a local fake — real screen capture, everything else simulated.
npm install
VITE_DEMO_MODE=true VITE_ADMIN_PASSWORD=dev npm run dev
# candidate: http://localhost:5173/ (take the exam)
# admin: http://localhost:5173/admin (watch + review — unlock with: dev)
# invigilator: http://localhost:5173/invigilatorOpen it in a Chromium-based browser (Chrome / Edge). The full local-dev guide,
including running against a real backend, is in LOCAL_DEV.md.
When you're ready to run it for real, the from-scratch Google Cloud deploy is in
docs/DEPLOY.md.
Every screenshot uses fabricated personas — this public repository contains no real candidate data.
| Watch the room live | Ranked results with integrity |
|---|---|
 |
 |
| Review the recording | The hire recommendation |
|---|---|
 |
 |
- Run it locally →
LOCAL_DEV.md - Deploy it for real →
docs/DEPLOY.md - Run an exam day →
docs/EXAM-DAY-OPS.md(operators) ·docs/CONDUCTOR-GUIDE.md(on-site room staff) - Understand how it all works →
docs/— the full, code-verified documentation set (architecture, every feature, the integrity + talent engine, the API, and the threat-model research behind the design) - Contribute →
CONTRIBUTING.md· Report a vulnerability →SECURITY.md
Three surfaces — candidate, admin, invigilator — in one self-hosted React app,
backed by a Node service on Google Cloud Run with Firestore + Cloud Storage, real
code execution via Judge0, and a separate evaluation service for the integrity +
talent engine. The full picture is in
docs/features/architecture-overview.md.
Released under the MIT License. Copyright (c) 2026 Aerele.