PlacePrep — AI-Powered College Placement Platform

Problem Statement
9 Practices
Practice1
Dashboard
Practice2
Practice3
Practice4
Practice5
Practice6
Practice7
Practice8
Practice9
Practice10
Practice11
Practice12
Practice13
Practice14

Inspiration

What it does

How we built it## Inspiration

Every year, millions of engineering students graduate in India unprepared for placement season. Placement cells run on spreadsheets, WhatsApp groups, and manual tracking. Students don't know where they stand. HODs can't report outcomes. Employers get unscreened candidates.

After 30 years in enterprise IT — hiring and managing thousands of professionals at MCIWorldCom, Alcatel, British Telecom, HCL, and Infosys — I saw this gap from both sides. PlacePrep is built to close it.

What it does

PlacePrep is a B2B AI-powered placement platform for colleges and universities:

Students practice aptitude, technical skills, coding, and mock interviews — tracked across 9 modules with a live leaderboard
Placement Officers get a digital dashboard replacing manual processes — student progress, company matching, placement status in one view
HODs see institutional analytics, outcome reports, and employer connect
AI Grading (Gemini) evaluates assignment submissions, gives structured feedback, and scores communication quality in real time

How we built it

Frontend: Vanilla JS, Firebase Firestore for real-time data
AI Engine: Google Gemini API for assignment grading, feedback generation, and mock interview evaluation
Scoring: Composite ranking formula — Quality 40%, Breadth 25%, Volume 20%, Streak 10%, Assignments 5%
Anti-contamination: uid-namespaced localStorage keys with ownership stamps to prevent cross-user data issues on shared college devices
Cloud Functions: Dynamic AI config loaded from Firestore — handles Gemini API failures gracefully without breaking submission pipeline

Challenges we faced

Gemini API grading failed silently due to a Google Cloud project mismatch — fixed by making Cloud Functions read AI config dynamically from Firestore
Score persistence was fragile across sessions — replaced with a universal localStorage interceptor for reliable cross-module tracking
Shared devices in college labs caused cross-user data contamination — solved with ownership-stamped uid-namespaced storage keys

What we learned

Gemini's structured output capability is powerful for grading open-ended answers — it can evaluate relevance, clarity, and structure simultaneously, which is exactly what placement interviews demand.