Case Study: Ladies Night Finder

The Challenge

Problem

Finding ladies night deals and promotions at bars and restaurants is fragmented — scattered across social media, venue websites, and word of mouth. Users needed a unified way to discover deals nearby, filtered by day and location.

Deal information changes frequently and is hard to verify
No single source of truth for venue promotions
Manual searches are time-consuming and unreliable
API costs can spiral quickly with naive implementations
Need to work across multiple cities and countries

The Solution

Solution

Built a sophisticated multi-stage search pipeline that combines AI services (Google Gemini, Perplexity Sonar), Google Places API, and intelligent caching. The system discovers, verifies, and serves venue deals with 70% API cost reduction through smart caching strategies.

System Architecture

Client

Mobile Apps

→

API Gateway

Express.js

→

AI Services
Gemini + Sonar

→

Data Layer

PostGIS + Redis

6-Stage Search Pipeline

0

Check Daily Venue Cache

Returns cached data if fresh, otherwise continues to next stage

1

Load Venue Intelligence Hints

Instagram-sourced venue data provides context for AI discovery

2

Gemini AI Venue Discovery

Uses intelligence hints to discover venues with deals

3

Temporal Processing

Day-of-week filtering, event scheduling, timezone handling

4

Store & Cache Verified Venues

Persist to PostgreSQL with PostGIS, 24-hour TTL caching

5

Merge, Deduplicate & Enrich

Include verified venues within 75km, remove duplicates, add deal details

Technology Stack

Component	Technology
Runtime	Node.js 18+
Framework	Express.js
Database	PostgreSQL 14+ with PostGIS
Cache	Redis
AI Services	Google Gemini 2.5, Perplexity Sonar
Location	Google Places API, OpenStreetMap Nominatim
Security	Helmet.js, express-validator, bcrypt
Mobile	React Native, Android (Kotlin), Expo

Node.js Express PostgreSQL PostGIS Redis Google Gemini Perplexity Sonar React Native Kotlin

Cost Optimization

API costs can quickly spiral with venue discovery apps. We implemented aggressive optimization strategies:

          
          
          
          Daily API Budget
        

┌────────────────────────┬─────────────┬───────────┐
│ Service                │ Daily Limit │ Est. Cost │
├────────────────────────┼─────────────┼───────────┤
│ Google Places Search   │ 500 calls   │ $16.00    │
│ Google Places Details  │ 300 calls   │ $5.10     │
│ Google Gemini          │ Unlimited   │ ~$0.50    │
│ Perplexity Sonar       │ Variable    │ ~$2.00    │
├────────────────────────┼─────────────┼───────────┤
│ Total                  │             │ ~$23.60   │
└────────────────────────┴─────────────┴───────────┘
        

        
          Optimization Strategies
        
        Aggressive Caching: 24-hour daily cache for processed venues, Redis for frequent lookups
API Call Batching: Max 2 API calls per venue verification
Intelligence Hints: Pre-scraped data reduces AI tokens needed
Distance Filtering: Venues >100km filtered out, 75km default radius

Results

70%

API cost reduction

<2s

Average response time

Global

Multi-city coverage

The system successfully serves real-time venue discovery across multiple cities with native mobile apps on both Android and iOS platforms.

Ladies Night Finder