Referral Copilot — AI-Powered LinkedIn Referral Message Generator

The Operational Problem

Job seekers spend countless hours writing awkward, boilerplate cold outreach messages on LinkedIn or sending generic copy that leads to near-zero response rates. Asking for referrals effectively requires precise role targeting, concise value propositions, and polite calls-to-action.

The Architectural Solution

Referral Copilot analyzes the target role, company name, candidate years of experience, core technical skills, and college/alumni background to synthesize high-reply LinkedIn referral requests and follow-up sequences in seconds. The system features a responsive form, real-time message previews, and instant clipboard copying.

• ✓AI-Powered Personalization: Crafts tailored referral messages incorporating role, company, skills, and alumni connection.
• ✓Follow-Up Sequence Generator: Creates polite follow-up messages referencing previous outreach.
• ✓Pre-written Template Fallbacks: Instant fallback to high-conversion curated templates if AI quotas or networks drop.
• ✓Multiple Monetization Tiers: Free daily tier, Starter credit packs, and Unlimited Pro monthly subscriptions.
• ✓Responsive Single-Page UI: Zero clutter, mobile-optimized interface built with React 19.
• ✓Secure Payment Verification: Razorpay checkout integration with server-side HMAC-SHA256 signature verification.

Technical Anatomy & Data Flow

Built as a decoupled client-server web application. The frontend is a React 19 single-page app utilizing responsive state management and copy utilities. The backend is an Express 5 service on Node.js with custom middleware for daily quota enforcement, validation, and request logging. Google Gemini Flash handles intelligent synthesis, while MongoDB (Mongoose 8) maintains user accounts, quota counters, and payment records.

// Deterministic prompt structure enforcing concise, high-reply outreach
const systemPrompt = `You are an expert career networking advisor. 
Generate a professional, personalized LinkedIn referral request message under 150 words.
Rules:
1. Mention the specific target role and company.
2. Highlight 1-2 core skills from candidate background.
3. If alumni/college is provided, establish genuine collegiate connection.
4. Keep the tone humble, appreciative, and low-pressure.
5. End with a simple, clear call to action.`;

const response = await geminiService.generateMessage({
  systemPrompt,
  userProfile: { company, role, skills, experienceYears, college },
  temperature: 0.4
});

Core Technology Stack

Security, Governance & Hardening

Strict daily rate limiting per IP and authenticated user ID, input sanitization preventing prompt injection attacks, server-side HMAC-SHA256 verification of Razorpay webhook signatures, and zero storage of sensitive third-party account credentials.