CampusLoop — Scalable Architecture for a Verified Multi-College Social Intranet

1. The Core Problem: Why College Social Fails

Higher education campuses in India (IITs, NITs, BIT Mesra, State Universities) lack a private, authentic digital commons. Student communication is currently fragmented across three broken channels:

• WhatsApp / Telegram Groups: Zero privacy, strict member limits, high admin burnout, and phone numbers exposed to strangers.
• Reddit / Instagram: No institutional boundary. Outsiders, spammers, and coaching institutes flood campus subreddits and comment sections.
• Toxic Anonymous Apps (e.g. Fizz / Sidechat clones): Lack geographic or email verification in India, leading to uncontrolled trolling and low student trust.

The Solution: CampusLoop solves this by enforcing Institutional Email Domain Gating paired with Cryptographically Decoupled Pseudonymity.

2. Cryptographic Pseudonymity & Auth Handshake

The primary technical challenge was achieving a dual guarantee: (1) verify that every user is an active student of that specific college, while (2) making it computationally impossible for anyone (including database administrators) to tie an anonymous confession back to a student's real identity.

┌────────────────────────────────┐       ┌────────────────────────────────┐
│ Student Email (abc@bitmesra.ac)│ ────► │  Institutional Domain Checker  │
└────────────────────────────────┘       └────────────────────────────────┘
                                                         │
                                                         ▼
                                         ┌────────────────────────────────┐
                                         │  Verified User Record (Auth)   │
                                         └────────────────────────────────┘
                                                         │
                                  One-Way Cryptographic Salt (HMAC-SHA256)
                                                         │
                                                         ▼
                                         ┌────────────────────────────────┐
                                         │ Pseudonym: "CosmicFalcon_26"   │
                                         │ (Feed / Polls / Confessions)   │
                                         └────────────────────────────────┘

• Domain Verification Engine: Validates against an indexed database of 1,350+ accredited Indian educational domains (`.edu.in`, `.ac.in`, college subdomains).
• Ephemeral Token Rotation: Session pseudonyms rotate periodically per category to eliminate cross-post linguistic fingerprinting.

3. Core Subsystems & Engineering Features

4. Multi-Tenant PostgreSQL Schema Design

Every data entity is strictly partitioned by college_id to guarantee zero cross-tenant contamination:

-- Multi-tenant Feed Post Schema with Isolation
CREATE TABLE campus_posts (
  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
  college_id UUID NOT NULL REFERENCES colleges(id) ON DELETE CASCADE,
  author_pseudo_id VARCHAR(64) NOT NULL,
  category VARCHAR(32) NOT NULL, -- 'confession', 'academic', 'hostel', 'lost_found'
  content TEXT NOT NULL,
  upvotes INT DEFAULT 0,
  comment_count INT DEFAULT 0,
  created_at TIMESTAMPTZ DEFAULT NOW()
);

CREATE INDEX idx_campus_posts_feed 
ON campus_posts (college_id, category, created_at DESC);

5. Hardest Concurrency Bugs & Performance Fixes

• WebSocket Reconnection Storms: When college campus WiFi dropped during major events, thousands of mobile clients reconnected simultaneously, overwhelming API endpoints. Fix: Implemented jittered exponential backoff on client PWA reconnects with connection pooling on the gateway.
• Race Conditions in High-Speed Campus Polls: Multiple simultaneous votes on trending campus elections caused negative delta counters. Fix: Replaced client-side state mutation with atomic PostgreSQL conditional upserts: UPDATE polls SET votes = votes + 1 WHERE id = $1 AND NOT ($2 = ANY(voter_hashes)).

6. Summary & Impact