We are building a Python-based backend that integrates with a MetaTrader 5 Expert Advisor (EA).
The EA sends structured market data (pair, session, ATR, volume, previous highs/lows, etc.) to a FastAPI server, and the server responds with AI-driven trade recommendations and parameters.
This backend will act as the “brain” of the EA, managing all decision logic, learning models, and data storage.
This is a serious build, not a small script — we’re creating a scalable, production-ready AI trading infrastructure.
Develop a FastAPI backend that receives JSON data from MT5 and returns trading bias + risk parameters.
Build an internal data processing pipeline that:
Cleans, normalizes, and stores incoming candle/trade data.
Appends trade outcomes to database for continuous retraining.
Integrate AI/ML logic to analyze new inputs and determine optimal strategy parameters for each session.
Support multiple currency pairs and sessions (London, New York).
Host a simple admin panel or API key system to manage user access and licenses.
Language: Python 3.10+
Framework: FastAPI (preferred)
Database: MySQL / PostgreSQL
Integration: Must handle HTTPS POST requests from EA (JSON format)
AI Component: Use historical data + outcome labels for retraining (e.g., LightGBM, CatBoost, or TensorFlow).
Data I/O: Must store candle data (OHLCV), ATR, session time, and order outcome results.
Scalability: Ability to serve multiple EA instances simultaneously with low latency (<200 ms).
Security: API key authentication per user.
EA sends a JSON packet:
Backend model interprets data → selects the most probable setup for that session.
Server returns:
{ "direction": "buy", "risk_percent": 0.03, "tp_rr": 4.0, "entry_zone": "1.0705-1.0715", "confidence": 0.82 }
EA executes trade and logs results → sent back to the API for retraining.
FastAPI backend with POST/GET endpoints.
Database schema for storing trades and session data.
Sample AI model integration (can start with basic logic; we’ll scale).
Log and retrain function (auto-update nightly).
Clear documentation and code comments.
Basic Docker or deployment guide for hosting (optional bonus).
✅ Python (FastAPI, Pandas, Numpy)
✅ MySQL or PostgreSQL (data storage and trade logs)
✅ Data mining and normalization (OHLCV, indicators)
✅ REST API design (JSON, authentication)
✅ Basic ML integration (LightGBM, Scikit-learn, or TensorFlow)
✅ Forex knowledge (candles, ATR, session logic)
Budget: $1,200 – $2,000 (depending on experience and performance)
Timeline: 2–3 weeks
Long-term work available (model retraining, scaling, and dashboard builds).
This project will directly power a commercial trading system.
Code must be modular, clean, and extensible — no shortcuts or unstructured scripts.
You’ll collaborate briefly with our MQL5 EA developer, so communication and clear handoff are important.
Please include:
Examples of APIs or trading data pipelines you’ve built (FastAPI, Flask, Django, etc.).
A short explanation of how you’d structure model retraining and data storage.
Availability and estimated completion time.