Home / Projects / Skillquest IFA
Enterprise · Case Study
Skillquest IFA
A gamified pre-screening platform that ranks candidates on how they actually solve problems — four timed cognitive puzzles instead of a résumé. Built for Insight Fusion Analytics.
React 18
TypeScript
Vite
Node.js + Express
MongoDB
OpenAI
Overview
Skillquest is a pre-screening assessment platform built for Insight Fusion Analytics. Instead of filtering applicants on résumés and CGPA, every candidate plays the same sequence of four timed puzzle games, and recruiters see a ranked dashboard of who solved the most, fastest. It's a first-round filter designed to surface how people think — not how well they wrote a CV.
The Problem It Solves
Early-stage and campus hiring leans on résumés, college names, and CGPA — weak, noisy proxies for actual problem-solving ability, and easy to inflate. The people hurt most are strong candidates without a polished résumé or a brand-name college, who get filtered out before anyone observes how they reason.
Skillquest's bet: put every applicant through identical, standardized, timed cognitive puzzles and rank them on observed performance — puzzles solved and time taken — rather than self-reported credentials. That gives recruiters a comparable, effort-based signal at the very top of the funnel, where the noise is worst.
Key Features
- Four hand-built cognitive games, each mapped to a named skill: Unblock Me (spatial reasoning & planning), Minesweeper (risk assessment & deductive logic), Water Capacity (logical sequencing & optimization), and a Question quiz (knowledge & recall) — all real engines, including ~100 authored Water-Capacity puzzles and difficulty-tagged Unblock levels.
- Sequential, timed assessment — each game unlocks only after the previous is submitted, with a 5-minute timer per game and auto-submit on expiry, so every candidate faces an identical, comparable path.
- Anti-cheating — tab-switch and window-blur detection with a three-strike warn-then-disqualify rule and a seven-day replay lock on failed games; a low-stakes practice mode per game keeps the real attempt fair.
- Weighted scoring & live leaderboard — a defined per-game scoring formula rolls up to a total, and a real-time leaderboard ranks completed assessments by score, tie-broken by total time.
- Recruiter dashboard — a candidate table with per-game drill-down (time spent, puzzles/levels completed), sortable and filterable by score, date, and role interest.
- AI-assisted question bank — admins upload existing PDF/DOCX/XLSX material and OpenAI extracts multiple-choice questions, with a rule-based fallback when no key is configured.
Product Decisions That Shaped It
- Standardized, effort-visible signal over self-report. Everyone faces the same timed puzzles; the recruiter's primary sort is observed performance, which is far harder to fake than a résumé line — and each game is explicitly tied to a cognitive skill so scores mean something.
- Bias reduction, framed honestly. Identical tasks and a performance-ranked leaderboard reduce the weight of pedigree at the ranking step. It mitigates credential bias rather than eliminating it — college and CGPA are still captured and shown to recruiters as supplementary context, not hidden.
- Make a screening step feel worth doing. Gamified framing, progress cues, a per-game practice mode, and a seven-day retry window are all there to keep candidates engaged instead of treating assessment as a chore.
- Sequential gating for fairness. Locking games in order prevents cherry-picking the easiest one and enforces one consistent path for apples-to-apples comparison.
- AI where it saves work, not where it's load-bearing. OpenAI turns documents recruiters already have into a question bank, but the assessment itself never depends on a model call — with a graceful no-key fallback.
Architecture
- Frontend: React 18 + TypeScript on Vite, a SPA with Tailwind, Framer Motion, and Radix primitives; the four game engines are substantial hand-written implementations, not off-the-shelf widgets.
- Backend: Node.js + Express 5 (ES modules) in a routes → controllers → models structure, with a CORS allowlist, rate-limited upload routes, and a global error handler.
- Data: MongoDB Atlas via Mongoose 8 — users, profiles, assessments, and the question bank.
- Auth & AI: Google OAuth with applicant/admin roles; OpenAI for question extraction from uploaded documents; deployed as separate frontend and backend on Vercel.
My Role & Engineering Highlights
- Built the platform end to end — four game engines, the assessment flow, the recruiter dashboard, and the Express/MongoDB backend.
- Authored the puzzle content and difficulty metadata (optimal-move tagging on Unblock levels, ~100 Water-Capacity puzzles) so scores reflect genuine difficulty.
- Designed the sequential-unlock + timer + anti-cheat flow to keep every candidate's attempt standardized and comparable.
- Wired an AI document-to-question-bank pipeline with a rule-based fallback, so recruiters can stand up a quiz from material they already have.
What I'd Harden Next
The honest engineering roadmap: move the trust boundary server-side. Scoring and the OAuth check currently run in the browser — the clear next steps are server-side score validation and server-side token verification so results can't be tampered with, plus real recruiter messaging (currently stubbed) and CSV export. Naming this openly is part of the case study: I know exactly where the product needs to mature.
Outcome
Skillquest reframes first-round hiring around a simple idea: rank people on how they solve problems, not on how they describe themselves. Building it meant designing fair assessment mechanics, four real game engines, and a recruiter workflow — the core problems of any HR-tech product.
Built for Insight Fusion Analytics as their gamified pre-screening platform, it's my deepest full-stack build — game logic, an Express/MongoDB backend, AI-assisted tooling, and a candidate experience designed to be fair and engaging at the top of the funnel.
Explore Related Projects