Campus Loop

The AI-powered, verified peer-to-peer student marketplace. Buy smarter, sell easier, study better - all within your own campus.

Comment

Campus Loop - Peer-to-Peer Campus Marketplace & Hub

A verified, hyper-local, AI-powered peer-to-peer student marketplace built exclusively for college campuses.

💡 Inspiration

Every semester, students across Indian college campuses repeat the same frustrating cycle: purchasing brand-new textbooks, lab equipment, and hostel essentials—items their seniors already own and would happily sell for a fraction of the price.

The current "solutions" are chaotic WhatsApp groups filled with unverified strangers, leading to zero accountability, spam, and safety concerns. We built Campus Loop because a student marketplace should feel as trustworthy as buying from a friend—not as sketchy as a random DM from an unknown number.

🏗️ What We Built

Campus Loop is a secured student marketplace where users can only join with a verified college email domain, listings are campus-scoped so buyers and sellers are always in the same physical location, and every transaction runs through a secure escrow system.

🔐 Verified Student Onboarding

  • Domain Whitelisting: Google Sign-In with automatic domain detection matching a whitelisted database table (institutes).
  • Instant Verification: Students signing up with official domains (e.g., @smail.iitm.ac.in, @iitb.ac.in) are auto-verified instantly.
  • Special Bypass Access: Hardcoded secure bypass endpoints allowing hackathon judges to verify instantly using test emails and a master bypass password.

🛒 Campus-Scoped Marketplace

  • Campus Scoping: Listings are automatically pre-filtered to the user's active campus—eliminating shipping logistics and keeping meetups on-campus.
  • Advanced Filters: Full text search, category filters (Books, Electronics, Lab Gear, Hostel, Cycles, Clothing), condition checks, price range sliders, and multi-parameter sort controls.

🤖 AI-Powered Selling Tools

We built three cutting-edge AI features directly into the product listing flow:

  1. AI Title & Description Generator: Sellers upload a photo, and the Groq Llama-4 Scout 17B Vision model analyzes the image to auto-fill a catchy product title and a conversion-optimized description.
  2. Smart Price Recommendation Engine: Powered by Google Gemini 1.5 Flash with real-time Google Search grounding. It cross-references the item title and condition against active Indian market rates and recommends an optimal price, minimum floor, and structured reasoning.
  3. AI Image Enhancer (Two-Phase Gemini Pipeline):
    • Phase 1 (gemini-3.1-flash-lite): Analyzes the uploaded photo (color tone, camera angle, lighting direction) and generates a custom background prompt optimized for maximum visual contrast.
    • Phase 2 (gemini-2.5-flash-image via Vertex AI): Swaps the messy dorm room background with a clean, studio-lit surface, keeping the product 100% unaltered.

💸 Escrow + Multi-Step Handover

  • Escrow Hold: Buyer pays → funds are held securely in platform escrow → order status transitions to PAID.
  • OTP Handover: At the physical meetup, the buyer inspects the product and provides a 6-digit OTP.
  • Brute-Force Protection: Sellers are locked out after 3 failed OTP attempts to prevent hacking or guessing.
  • Seller Payout Form: Verifying the OTP updates the status to otp_verified (completing the transaction on the buyer's side). The seller is then prompted to enter their bank details (Holder Name, Bank Name, Account Number, IFSC) to transfer the funds and mark the order as completed.
  • Escrow Guarantee: Zero platform fees; escrow coverage is fully included.

💬 Secure Transaction Chat

  • Real-Time Messaging: Built-in chat channel for coordination.
  • PII Redaction Filter: Powered by Groq Llama 3.1 8B. Intercepts outgoing messages, automatically redacting phone numbers, emails, and social handles with [REDACTED] to keep transactions safe and on-platform.
  • Location Sharing: Users share GPS coordinates or select campus landmark pins, rendering as interactive Google Maps iframes directly inside chat bubbles.

📚 Study Hub & Student Directories

  • Academic Sharing: Students upload and share notes, lecture materials, and exam papers.
  • S3 + CDN: Files are streamed directly to Amazon S3 and served via CloudFront CDN.
  • Events & Clubs: Campus-scoped events list with RSVP member counts and student clubs directory with WhatsApp group invite links.

📱 PWA Support

  • Installable on Android and iOS directly from the browser.
  • Service worker caching provides offline availability of resources.

🟠 How We Used AWS

AWS is the backbone of Campus Loop's data infrastructure:

  • Amazon Aurora PostgreSQL (Serverless v2): Our core relational database hosting users, listings, orders, chat messages, and verification queues.
    • Scale-to-Zero: Auto-scales to 0 ACUs when idle—minimizing costs when inactive.
    • Instant Scaling: Instantly scales up to 2 ACUs under query loads with zero latency.
  • Amazon S3: Product images and Study Hub PDFs are uploaded directly via multipart FormData streams.
  • Amazon CloudFront CDN: Caches and distributes media assets at edge locations globally to ensure ultra-low latency image rendering.

⚙️ Technical Architecture

Technical Architecture

💰 Monetization Model

Campus Loop operates on a two-stage monetization roadmap:

Stage 1: Active User Acquisition (Current Stage)

  • Automated Amazon Affiliate Links: Used book/textbook listings run through an NLP analyzer that automatically injects a "Buy New on Amazon" button containing our affiliate tag (campusloop-21).
  • Google AdSense: Integrated ad banners in the Study Hub resource library and listing feeds.

Stage 2: Scale Operations

  • Transaction Fee: A 1-2% platform fee on completed escrow handovers.
  • Promoted Listings: Micro-fees paid by sellers to pin their listings to the top of category feeds.
  • Brand Partnerships: Direct listing channels for local vendors and campus brands.

🚧 Challenges We Faced

  1. Database Migration Under Deadline: Porting our relational schema from Supabase to Aurora PostgreSQL serverless required configuring native connection pooling (pg Pool) and rewriting queries to match Aurora's SSL parameters.
  2. Double-Phase AI Image Pipeline: Designing the background replacer required tuning Phase 1 prompts on gemini-3.1-flash-lite to ensure the generated background prompt provided high contrast (e.g. dark product -> light background) instead of matching tones.
  3. Imagen 3 Model Deprecation: The sudden deprecation announcement of imagen-3.0 forced us to pivot our background-swapping editor. We successfully rebuilt the Vertex AI pipeline using the new gemini-2.5-flash-image model via the Google Gen AI SDK.
  4. Real-Time PII Filters: Running LLM chat filtering on every keypress caused message lag. We resolved this by failing open to a regex fallback if the LLM filter took longer than 800ms.

📖 What I Learned

  • Aurora Serverless v2 is highly cost-efficient; compute costs remained under $50 for the entire hackathon period.
  • Multimodal LLM prompts can achieve production-quality image background editing when paired with structured contrast instructions.
  • Trust features (such as escrow, OTP locks, PII redactions, and seller verification) are essential to creating a secure campus marketplace.

Built With

  • amazon-associates
  • amazon-aurora-postgresql
  • amazon-cloudfront-cdn
  • amazon-web-services
  • firebase-authentication
  • gemini-2.5-flash-image
  • gemini-3.1-flash-lite
  • google-adsense
  • google-gemini-api
  • google-maps-embed-api
  • google-vertex-ai
  • groq
  • llama-3.1-8b
  • llama-4-scout-17b-vision
  • next.js
  • next.js-app-router
  • node.js
  • pg-(node-postgres)
  • pwa
  • react
  • react-19
  • resend
  • service-worker
  • tailwind-css-v4
  • tanstack-query
  • turbopack
  • typescript
  • upstash-redis
  • vercel
  • zustand
Share this project:

Updates