GridDB Logo
GridDB Developers
Download
Back to Gallery

Richie Rich

By Tejas Hari, Saatvik Sinha, Rahul Raman Finance
Richie Rich
Finance

Richie Rich

Intelligent Multi-Asset Portfolio Tracker

Download the Richie Rich Presentation

Elevator Pitch

Richie Rich is a real-time portfolio management dashboard that tracks stocks, crypto, and commodities across global markets in a single unified interface. Built on GridDB’s time-series capabilities, it stores granular price data, generates machine-learning-driven insights, and automates weekly performance reports using GitHub Actions.

About the Project

Motivation

Modern hedge funds manage portfolios spanning:

  • Global equities
  • Cryptocurrencies
  • Commodities
  • Derivatives

However, today’s tooling is fragmented:

  • Bloomberg terminals for US equities
  • Specialized NSE tools for Indian markets
  • Separate crypto exchange dashboards
  • Independent commodity tracking platforms

This fragmentation introduces information latency and increases operational overhead.

Richie Rich was built as a unified platform that:

  • Aggregates real-time multi-asset data
  • Stores granular historical price movements
  • Powers predictive models to identify alpha-generating opportunities

Traditional relational databases struggle with high-frequency financial tick data. GridDB’s time-series architecture provides optimized write performance and queryable history, enabling price updates every 15–20 seconds without bottlenecks.

Technical Architecture

System Components

1. Backend Layer

  • FastAPI server orchestrating data ingestion
  • Sources:
    • CoinGecko API (cryptocurrency prices)
    • Simulated models (equities)
  • Async price fetching with exponential backoff
  • Background worker persists snapshots to GridDB every 20 seconds when frontend activity is detected

2. Storage Layer

  • GridDB Cloud – TIME SERIES container
  • Optimized for chronological writes

Schema:

{ timestamp, symbol, price, change24h }

  • Row key collision handled via millisecond-offset timestamps for concurrent multi-symbol inserts

3. Presentation Layer

  • React 18 + TypeScript
  • Vite build tooling
  • Tailwind CSS styling

Portfolio Valuation Formula:

V_portfolio = Σ (Qi × Pi)

Where:

  • Qi = quantity held
  • Pi = current market price

Weighted 24h Performance:

Δ24h = (Σ (Qi × Pi × change_i,24h)) / V_portfolio

Machine Learning Integration

  • XGBoost classifier for buy / sell / hold recommendations
  • Feature engineering from GridDB historical data:

Features = { price, Δ24h, highest, lowest, range, range% }

  • Model trained on stockPrice.csv
  • Recommendations exposed via /recommendations API endpoint

Planned Enhancements:

  • Time-series forecasting models
  • Moving averages and correlation matrices using GridDB aggregations

Automation Infrastructure

A GitHub Actions workflow automates weekly portfolio exports:

  • Triggered via:
    • Cron (Sunday, 00:00 UTC)
    • Manual dispatch
  • Fetches portfolio state from backend
  • Exports timestamped CSV to /data
  • Commits and pushes via bot credentials
  • Artifacts retained for 90 days

This provides auditable performance history for compliance and backtesting.

Key Learnings

Time-Series Databases Matter

  • GridDB’s TIME SERIES containers align naturally with financial analysis
  • Traditional RDBMS indexing fails at scale for time-ordered inserts

Row Key Semantics

  • TIME SERIES containers use timestamps as row keys
  • Concurrent inserts caused overwrites
  • Solved via millisecond offsets per symbol

Real-Time API Limitations

  • CoinGecko rate limits
  • Yahoo Finance blocks automation
  • Required caching, fallback logic, and exponential backoff

Frontend Performance

  • Polling every 15 seconds stresses React state
  • Proper useEffect cleanup was critical to avoid memory leaks

Realistic Price Modeling

  • Initial static pricing produced no visible movement
  • Correct approach:

price_fluctuated = price_base × (1 + Δpct / 100)

GitHub Actions Permissions

  • Default GITHUB_TOKEN is read-only
  • Required:

permissions: contents: write

Development Timeline

Week 1:

  • FastAPI endpoints
  • React frontend with mock data

Week 2:

  • CoinGecko integration
  • Real-time valuation logic
  • Async background tasks

Week 3:

  • GridDB Cloud deployment
  • TIME SERIES container setup
  • Row key collision fixes

Week 4:

  • XGBoost ML model
  • Recommendation API
  • Glassmorphic dark UI

Week 5:

  • Price fluctuation bug fixes
  • Render + Vercel deployment
  • CORS configuration

Week 6:

  • GitHub Actions automation
  • Weekly CSV exports
  • Cron scheduling

Current System Features

  • Real-time tracking for 10 assets
  • Automatic GridDB persistence every 20 seconds
  • Portfolio valuation with weighted 24h performance
  • ML-powered buy/sell/hold recommendations
  • Weekly automated CSV exports
  • Responsive glassmorphic UI
  • REST API with CORS support
  • Environment-based configuration

Technical Challenges & Solutions

GridDB Authentication

  • Requires Base64-encoded Basic Auth header
  • Exact formatting needed to avoid 401 errors

TIME SERIES Row Key Collisions

  • Solved using millisecond offsets per symbol

API Rate Limiting

  • CoinGecko free tier: 50 calls/min
  • Implemented exponential backoff + caching

React Interval Bugs

  • Improper cleanup caused overlapping fetches
  • Fixed with strict dependency management

GitHub Actions Issues

  • Workflow location requirements
  • Token permissions
  • Deprecated actions upgraded

Deployment Connectivity

  • Render free tier uses dynamic IPs
  • GridDB Cloud free tier requires static IPs
  • Documented as a known limitation

Production Solutions:

  • Paid GridDB tier
  • Static IP proxy
  • Same-VPC deployment

Technology Stack

Backend

  • Python 3.11
  • FastAPI
  • Uvicorn

Frontend

  • React 18
  • TypeScript
  • Vite
  • Tailwind CSS

Database

  • GridDB Cloud (TIME SERIES)

Machine Learning

  • XGBoost
  • pandas
  • numpy

APIs

  • CoinGecko
  • GridDB WebAPI
  • Simulated equity models

DevOps

  • GitHub Actions
  • Render (backend)
  • Vercel (frontend)

Summary

Richie Rich demonstrates how a dedicated time-series database like GridDB enables high-frequency financial data ingestion, real-time analytics, and machine-learning-driven insights. By unifying multi-asset tracking into a single platform and automating reporting pipelines, the project showcases a scalable, production-aware approach to modern portfolio management.

Team Members: Tejas Hari, Saatvik Sinha, Rahul Raman