About

This canvas implements a production crypto trading signal engine that fetches 1-minute futures data for 8 coins (BTC, ETH, SOL, BNB, XRP, DOGE, NEIRO, ZEREBRO), engineers 14 technical indicators across a parallel fleet, labels targets using 5-bar forward returns with per-coin thresholds, trains per-coin MLP classifiers with hyperparameters tuned for volatility levels, and emits real-time predictions gated by per-coin confidence thresholds and ATR(14) volatility warnings to Telegram. The workflow orchestrates data ingestion → feature engineering (fleet) → aggregation → labeling → model training with RL reward simulation → live inference with multi-gate filtering and alerting.