InnovateX United
AI POWERED TRADING TERMINAL
"Your partner in financial growth."
Submitted
May 22, 2026, 8:55 AM
Last Updated
May 22, 2026, 8:55 AM
Project Links
Special Mention
This project was recognized for outstanding achievement in the Fintech category at Catalyst 2K26.
Problem Statement
Legacy trading terminals suffer from severe UI latency due to JSON parsing overhead and bloated Electron architectures that crash under high data volume. Furthermore, traders face decision paralysis from fragmented manual analysis of charts and news, while relying on highly insecure .env configurations. These legacy systems are too slow, resource-heavy, disconnected, and insecure for modern high-frequency quantitative execution
Solution & AI Usage
In our project, AI eliminates human decision paralysis by acting as a fleet of specialized, real-time microservices: Sentiment Analysis: Claude 3 Haiku reads breaking financial news, converting unstructured text into low-latency quantitative sentiment scores. Anomaly Detection: DeepSeek V4 Pro (via NVIDIA NIM) analyzes raw order flow vectors to intercept hidden institutional liquidity imbalances and volume anomalies. Predictive Modeling: A Rust-native OLS regression engine mathematically forecasts probabilistic price trajectories (the "Ghost Line"). Consensus Resolution: A central Rust Aggregator fuses these AI outputs with deterministic technical patterns, automatically resolving data conflicts to generate singular, high-confidence trading directives.
Full Description
Our project is an institutional-grade, AI-driven quantitative trading desktop application engineered to eliminate UI latency and human decision paralysis. Built on a zero-overhead Rust and Tauri container, the platform operates at 0% idle CPU, utilizing strict client-side Next.js rendering to prevent crashes. System Data Flow: Data Ingestion: A headless Rust daemon captures live market data via broker WebSockets, dynamically streaming raw ticks into an Apache Kafka event cluster and a high-throughput QuestDB time-series database. AI & Quant Processing: Decentralized microservices analyze the Kafka streams in real-time. A Node.js agent utilizes Claude 3 Haiku for instant news sentiment. A DeepSeek V4 model detects institutional volume anomalies. Meanwhile, a Rust mathematical engine calculates probabilistic price trajectories using OLS linear regression (the "Ghost Line"). Consensus Aggregation: A central Rust engine fuses these predictive AI outputs with deterministic technical patterns (like ORB and candlestick breakouts), automatically resolving data conflicts to generate unified, actionable execution signals. Zero-Latency Visualization: Bypassing traditional JSON bloat, the backend streams processed data to the frontend via raw binary buffers (bincode). The React UI instantly renders updates on TradingView charts, while all credentials remain secured in a local cryptographic vault.
Tech Stack
Screenshots
4 attached
