diff --git a/.jules/bolt.md b/.jules/bolt.md
index 7e410c4..43ca330 100644
--- a/.jules/bolt.md
+++ b/.jules/bolt.md
@@ -9,3 +9,7 @@
 ## 2026-06-10 - [Lazy Loading Scikit-learn & NumPy]
 **Learning:** Top-level imports of `numpy` and `joblib` plus loading a serialized model file (`.joblib`) during FastAPI startup adds significant overhead (e.g., ~3 seconds). Refactoring this into a service with `cached_property` and local imports deferred the cost until the first request.
 **Action:** Move all heavy ML model loading and their dependencies into lazy-loaded properties to ensure near-instant application startup.
+
+## 2026-06-20 - [ML Prediction Hot Path Optimization]
+**Learning:** Passing a list of lists directly to Scikit-learn's `model.predict()` avoids the significant overhead of `numpy` array creation and reshaping for single-sample inference. Combining this with per-instance `lru_cache` (via a `cached_property` returning a closure) provides safe, massive performance wins for repeat requests without memory leaks or unhashable `self` issues.
+**Action:** For single-sample ML inference, prefer native Python structures over `numpy` when supported by the model, and use per-instance memoization closures to optimize hot paths safely.
diff --git a/ml_service.py b/ml_service.py
index bee0bde..c5d2bb3 100644
--- a/ml_service.py
+++ b/ml_service.py
@@ -15,12 +15,25 @@ def model(self):
             logger.error(f"Error loading ML model: {e}")
             return None
 
+    @cached_property
+    def _predict_internal(self):
+        """Internal method to provide instance-level caching for predictions."""
+        from functools import lru_cache
+
+        @lru_cache(maxsize=128)
+        def _cached_predict(features_tuple):
+            # Scikit-learn models can often take a list of lists directly,
+            # avoiding the overhead of numpy array creation for a single sample.
+            prediction = self.model.predict([features_tuple])
+            return int(prediction[0])
+
+        return _cached_predict
+
     def predict(self, features_list):
-        """Perform prediction using the lazy-loaded model."""
+        """Perform prediction using the lazy-loaded model with caching."""
         if self.model is None:
             raise RuntimeError("ML model could not be loaded")
 
-        import numpy as np
-        features = np.array(features_list).reshape(1, -1)
-        prediction = self.model.predict(features)
-        return int(prediction[0])
+        # Use a tuple to make features hashable for lru_cache
+        features_tuple = tuple(features_list)
+        return self._predict_internal(features_tuple)