Vibe Coding

Trading Analytics Platform — Institutional-Grade NSE System

Full-stack institutional-grade trading analytics system for Indian equity markets (NSE) built end-to-end with vibe coding. React Native mobile frontend with Bloomberg Terminal-style UI, Python Flask backend with 46+ service files, ML prediction engine using XGBoost with 80 technical indicators (95.7% win rate, Sharpe Ratio 15.17), and 6 MCP servers giving agents live access to a 459 MB unified trading database.

Client Personal Project
Delivered Jun 2025
Type Vibe Coding
Technologies 16 tools
Trading Analytics Platform — Institutional-Grade NSE System

Project Overview

A full-stack institutional-grade trading analytics system for Indian equity markets (NSE), built end-to-end as a personal project using vibe coding. The goal was to bring the depth of institutional trading tooling to an individual developer — from data collection through ML prediction to real-time order tracking — all in one cohesive system.

Frontend — React Native Mobile App

  • Bloomberg Terminal-style UI with real-time data streaming via WebSocket/Socket.IO.
  • Lightweight Charts integration for candlestick, line, and volume chart rendering.
  • Real-time order tracking panels, portfolio overview, and strategy performance dashboards.
  • Mobile-responsive across iOS and Android via Expo.

Backend — Python Flask (46+ service files)

  • Data collection: AngelOne SmartAPI integration, NSE scraper, and 8+ news sources aggregated in real time.
  • ML pipeline: XGBoost prediction engine with 80 technical indicators; validated via Monte Carlo simulation and Walk-Forward analysis, achieving a 95.7% win rate with a Sharpe Ratio of 15.17.
  • Backtesting engine: 4 trading strategies within a unified framework — ML Prediction, Trend Following, Mean Reversion, and Momentum — with position sizing and portfolio optimisation.
  • Risk management: Dynamic position sizing, max drawdown controls, and portfolio-level risk constraints.
  • Real-time order tracking: Live order status updates from AngelOne API with P&L computation.

AI & NLP Layer

  • FinBERT sentiment analysis — finance-domain NLP model aggregating sentiment from 8+ news sources in real time.
  • FII/DII institutional flow tracking from NSE data — correlating institutional activity with market movement.
  • TensorFlow / PyTorch models for supplementary signal generation.

MCP Integration

Integrated 6 MCP servers to give AI agents live access to the 459 MB unified trading database (1.99M rows, 16 data types) for real-time analysis and query assistance without hallucination:

  • DuckDB MCP, SQLite MCP, Prometheus MCP, Filesystem MCP, Git MCP, Memory MCP.

Database

  • SQLite for structured trade and order data; DuckDB for analytical workloads on the 1.99M-row unified trading database.

Technology Stack

React Native Expo Python Flask Flask-SocketIO XGBoost TensorFlow PyTorch FinBERT SQLite DuckDB AngelOne SmartAPI NSE Scraper WebSocket MCP Servers Lightweight Charts

Project Info

  • Client Personal Project
  • Delivered June 2025
  • Type Vibe Coding

Stack

React Native Expo Python Flask Flask-SocketIO XGBoost TensorFlow PyTorch FinBERT SQLite DuckDB AngelOne SmartAPI NSE Scraper WebSocket MCP Servers Lightweight Charts

Available for new projects

Let's build something great together

Open to freelance, contract, and full-time opportunities. Let's discuss your next project.