Credit Ninja
FinTech Search Platform · 2021
Screenshots
Overview
A large-scale platform that enables users to search and analyze credit data for 200,000+ companies. The platform serves financial professionals who need quick, reliable access to credit information for business decisions.
The Challenge
Situation
The client needed a way to make their massive database of company credit data accessible to users. Existing solutions were either too slow, too expensive, or lacked the search precision required for financial due diligence.
Task
Build a fast, secure search platform that could handle complex queries across 200,000+ company records while providing real-time results and maintaining strict data security.
My Approach
- Designed and implemented the core backend architecture from scratch, focusing on scalability and performance
- Built a high-performance search system using MongoDB Atlas Search with custom scoring and filtering
- Implemented secure OTP-based authentication to protect sensitive financial data
- Created a GraphQL API layer for flexible, efficient data fetching from the frontend
Results & Impact
- Search queries return results in under 500ms across the full 200,000+ company dataset
- Platform successfully launched and adopted by financial professionals
- Authentication system maintained zero security incidents
- Architecture scales efficiently as the database grows
Key Metrics
System Architecture
Dual-Search Architecture: MongoDB Atlas Search (primary queries) + Elasticsearch (autocomplete/aggregations) with 3-tier Redis caching
Performance: Sub-50ms autocomplete, 150-250ms uncached queries across 200K+ documents. 40% cache hit rate.
Technical Highlights
Single search engine couldn't provide both fuzzy matching and advanced faceted search at scale.
Combined MongoDB Atlas Search for primary queries with Elasticsearch for autocomplete and aggregations. Edge N-gram tokenizer for real-time suggestions.
Sub-50ms autocomplete, 150-250ms uncached queries across 200K+ documents.
Database load was unsustainable at 5000+ searches/minute peak traffic.
Implemented Apollo Client cache (5min), Redis cluster (1h search, 24h details), and MongoDB as source of truth with LRU eviction.
40% cache hit rate, 60% reduction in database costs, 70% faster average response.
Financial data required protection against brute-force, injection, and token theft attacks.
Granular rate limiting (Redis-backed), JWT with refresh token rotation, httpOnly SameSite cookies, MongoDB sanitization, GraphQL depth limiting.
Zero security breaches, 99%+ attack prevention, prevented 1M+ malicious requests/month.
Tech Stack
Need a similar search platform for your data?
Let's discuss your project over a free discovery call.
Book a Discovery Call