Trade VIsion

TradeVision: AI-powered crypto platform combining real-time analysis, RAG search, ML forecasts, and simulated blockchain trades for smarter, faster decisions.

Comment

Inspiration

The cryptocurrency market is overwhelming, with traders drowning in price charts, social sentiment, and news. I wanted to build a platform that consolidates all this data, leverages AI for actionable insights, and provides a risk-free environment to test trading strategies. The goal: smarter, faster, and more confident decisions.

What it does

TradeVision is an AI-powered crypto trading platform that:

  • Extracts and syncs live crypto data using a custom Fivetran connector.
  • Provides 7-day ML price forecasts using Prophet.
  • Delivers contextual market insights with Google Gemini AI.
  • Retrieves relevant news and reports using Elasticsearch + RAG.
  • Simulates blockchain trades using Hedera (mock mode) with instant transaction finality.

How we built it

  • Data Pipeline: CoinGecko API → Fivetran connector → BigQuery (ETL + historical data sync).
  • AI/ML: Gemini AI analyzes market and sentiment; Prophet predicts prices.
  • RAG Search: Elasticsearch indexes news and reports for hybrid BM25 + semantic search.
  • Blockchain Simulation: Hedera mock trades simulate real-world execution safely.
  • Frontend: Streamlit dashboard with interactive Plotly charts for visualization.

Challenges we ran into

  • Integrating multiple systems (Fivetran, BigQuery, Gemini AI, Elasticsearch, Hedera).
  • Maintaining data consistency during incremental updates.
  • Optimizing real-time performance with large historical datasets.
  • Simulating blockchain trades accurately with confirmations and balances.

Accomplishments that we're proud of

  • Built a production-ready platform deployed on Streamlit Cloud.
  • First platform combining RAG + Real-time AI + Blockchain simulation.
  • Custom Fivetran connector with 60-day incremental sync.
  • Transparent and automated pipeline, zero maintenance.

What we learned

  • Hands-on experience in data engineering, AI, ML, search, and blockchain integration.
  • How to combine structured and unstructured data for actionable insights.
  • Designing forecast confidence metrics for data-driven trading decisions.

[ \text{Trade Signal Confidence} = \frac{\text{Predicted Return}}{\text{Forecast Uncertainty}} ]

What's next for TradeVision

  • Support multiple exchanges (Binance, Coinbase, Kraken).
  • Implement advanced ML models (LSTM, Transformer) for better forecasts.
  • Add portfolio optimization algorithms and mobile app integration.
  • Deploy on Hedera mainnet for real trades and integrate regulatory dashboards.

Built With

Share this project:

Updates