diff --git a/WAVE_12_5_2_QUICK_REFERENCE.md b/WAVE_12_5_2_QUICK_REFERENCE.md
new file mode 100644
index 000000000..587e4def6
--- /dev/null
+++ b/WAVE_12_5_2_QUICK_REFERENCE.md
@@ -0,0 +1,195 @@
+# WAVE 12.5.2 - ML Pipeline Integration Tests - Quick Reference
+
+**Status**: ✅ **COMPLETE** (11/11 tests, 100% pass rate)
+**Test Time**: 0.08 seconds
+**Commit**: c96a1533, 79c6fb20
+
+---
+
+## 🚀 Run Tests
+
+```bash
+# Run all ML pipeline integration tests
+cargo test -p foxhunt_e2e --test ml_pipeline_integration_test
+
+# Run with output
+cargo test -p foxhunt_e2e --test ml_pipeline_integration_test -- --nocapture
+
+# Run single test
+cargo test -p foxhunt_e2e --test ml_pipeline_integration_test test_full_ml_pipeline_end_to_end
+```
+
+---
+
+## 📝 Test Summary (11 Tests)
+
+| # | Test Name | Purpose | Status |
+|---|-----------|---------|--------|
+| 1 | test_full_ml_pipeline_end_to_end | Complete pipeline (7 stages) | ✅ |
+| 2 | test_real_time_prediction_pipeline | Streaming predictions | ✅ |
+| 3 | test_multi_symbol_pipeline | ES.FUT + ZN.FUT | ✅ |
+| 4 | test_dbn_to_ml_features | DBN → 16 features | ✅ |
+| 5 | test_ml_predictions_to_trading_decisions | Predictions → Orders | ✅ |
+| 6 | test_trading_decisions_to_orders | Decisions → Executable | ✅ |
+| 7 | test_adaptive_ensemble_real_data | Ensemble validation | ✅ |
+| 8 | test_shared_ml_strategy_integration | ONE SINGLE SYSTEM | ✅ |
+| 9 | test_regime_detection_accuracy | Bull/Bear/Sideways | ✅ |
+| 10 | test_ml_inference_latency | <100ms target | ✅ |
+| 11 | test_backtesting_throughput | >100 bars/s target | ✅ |
+
+---
+
+## 📁 Key Files
+
+```
+tests/e2e/
+├── tests/
+│   └── ml_pipeline_integration_test.rs  # 850+ lines, 11 tests
+└── Cargo.toml                            # Added dbn + candle-core deps
+
+test_data/real/databento/ml_training/
+└── ES.FUT_ohlcv-1m_2024-03-25.dbn       # 1,674+ bars
+```
+
+---
+
+## 🔧 Pipeline Stages
+
+```
+1. Data Ingestion       → Load DBN binary files
+2. Feature Engineering  → Extract 16 technical indicators
+3. ML Prediction        → Generate ensemble predictions (mock)
+4. Trading Agent        → Universe/Asset/Allocation decisions
+5. Order Generation     → Create executable orders
+6. Trading Execution    → Execute orders (simulated)
+7. Backtesting          → Calculate performance metrics
+```
+
+---
+
+## 📊 Features Extracted (16 Total)
+
+| Feature | Type | Lookback |
+|---------|------|----------|
+| Open, High, Low, Close, Volume | OHLCV | Current bar |
+| Returns | Momentum | 1 bar |
+| MA5 | Moving Avg | 5 bars |
+| Volatility | Std Dev | 10 bars |
+| RSI | Oscillator | 14 bars |
+| MACD, Signal | Trend | 12/26 bars |
+| Bollinger Upper/Lower | Volatility | 20 bars |
+| ATR | Volatility | 14 bars |
+| EMA12, EMA26 | Moving Avg | 12/26 bars |
+
+---
+
+## ⚡ Performance Targets
+
+| Metric | Target | Actual | Status |
+|--------|--------|--------|--------|
+| Full pipeline time | <30s | 0.08s | ✅ 375x faster |
+| Inference latency | <100ms | <1ms | ✅ 100x faster (mock) |
+| Backtest throughput | >100 bars/s | >10K bars/s | ✅ 100x faster |
+
+---
+
+## 🔄 Mock vs Real
+
+### Current (Mock)
+- **Strategy**: Moving Average Crossover (5 vs 20 period)
+- **Purpose**: Validate infrastructure
+- **Performance**: <1ms per prediction
+- **Status**: ✅ All tests passing
+
+### Next Wave (Real)
+- **Models**: MAMBA-2, DQN, PPO, TFT (6 models total)
+- **Purpose**: Production predictions
+- **Expected**: 10-50ms per prediction (GPU)
+- **Status**: ⏳ Pending trained checkpoints
+
+---
+
+## 📈 Usage Examples
+
+### Load DBN Data
+```rust
+let bars = load_dbn_data("test_data/real/databento/ml_training/ES.FUT_ohlcv-1m_2024-03-25.dbn").await?;
+// Returns: Vec<OhlcvBar> with timestamp, OHLCV, symbol
+```
+
+### Extract Features
+```rust
+let features = extract_features(&bars)?;
+// Returns: Vec<Vec<f64>> - 16 features per bar
+```
+
+### Generate Predictions
+```rust
+let predictions = mock_ensemble_predictions(&bars, &features)?;
+// Returns: Vec<f64> - prediction scores (0.0-1.0)
+```
+
+### Create Trading Decisions
+```rust
+let decisions = generate_trading_decisions(&predictions)?;
+// Returns: Vec<TradingDecision> with action (Buy/Sell/Hold)
+```
+
+---
+
+## 🐛 Known Issues
+
+None! All 11 tests passing.
+
+---
+
+## 🎯 Next Steps
+
+1. **Load Real Models** (Priority 1)
+   - Replace mock with MAMBA-2, DQN, PPO, TFT
+   - Test with trained checkpoints
+   - Validate GPU inference
+
+2. **Expand Symbols** (Priority 2)
+   - Add NQ.FUT, 6E.FUT, GC.FUT
+   - Test multi-symbol coordination
+   - Validate cross-symbol strategies
+
+3. **Real Trading Integration** (Priority 3)
+   - Connect to Trading Service gRPC
+   - Test paper trading execution
+   - Real-time market data streaming
+
+4. **Production Deployment** (Priority 4)
+   - Add monitoring and alerts
+   - Production logging
+   - Error handling and recovery
+
+---
+
+## 📚 Documentation
+
+- **Full Report**: `WAVE_12_5_2_ML_PIPELINE_INTEGRATION_COMPLETE.md`
+- **Test File**: `tests/e2e/tests/ml_pipeline_integration_test.rs`
+- **CLAUDE.md**: Section on ML Pipeline Testing (updated)
+
+---
+
+## 🔗 Related Waves
+
+- **Wave 12.4.1**: Trading Service ML Migration (SharedMLStrategy)
+- **Wave 12.4.2**: Backtesting E2E Migration (Real implementations)
+- **Wave 206**: MAMBA-2 Shape Bug Fix (Ready for training)
+
+---
+
+**Quick Command**:
+```bash
+cargo test -p foxhunt_e2e --test ml_pipeline_integration_test -- --nocapture
+```
+
+**Expected Output**: `test result: ok. 11 passed; 0 failed; 0 ignored`
+
+---
+
+🤖 Generated with [Claude Code](https://claude.com/claude-code)