diff --git a/src/models/model_utils.py b/src/models/model_utils.py
index 8b13789..9ee4d84 100644
--- a/src/models/model_utils.py
+++ b/src/models/model_utils.py
@@ -1 +1,115 @@
+import joblib
+import os
+import json
+from datetime import datetime
 
+MODEL_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), "../../models_saved"))
+METADATA_FILE = os.path.join(MODEL_DIR, "metadata.json")
+
+
+def _load_metadata() -> dict:
+    """Load existing metadata registry or return empty dict."""
+    if os.path.exists(METADATA_FILE):
+        with open(METADATA_FILE, "r") as f:
+            return json.load(f)
+    return {}
+
+
+def _save_metadata(metadata: dict):
+    """Persist metadata registry to disk."""
+    os.makedirs(MODEL_DIR, exist_ok=True)
+    with open(METADATA_FILE, "w") as f:
+        json.dump(metadata, f, indent=4)
+
+
+def save_model(model, version_tag: str, hyperparams: dict = None, metrics: dict = None):
+    """
+    Save a trained model with a version tag and metadata.
+
+    Args:
+        model:         Trained sklearn model
+        version_tag:   e.g. 'v1_base', 'v2_tuned'
+        hyperparams:   Dict of hyperparameters used (optional)
+        metrics:       Dict of evaluation metrics (optional)
+
+    Example:
+        save_model(rf, "v1_base",
+                   hyperparams={"n_estimators": 100},
+                   metrics={"accuracy": 0.91})
+    """
+    os.makedirs(MODEL_DIR, exist_ok=True)
+
+    filename = f"model_{version_tag}.pkl"
+    path = os.path.join(MODEL_DIR, filename)
+    joblib.dump(model, path)
+
+    metadata = _load_metadata()
+    metadata[version_tag] = {
+        "filename": filename,
+        "training_date": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+        "hyperparameters": hyperparams or {},
+        "metrics": metrics or {},
+    }
+    _save_metadata(metadata)
+
+    print(f"[model_utils] Model saved as '{filename}'")
+    print(f"[model_utils] Metadata updated for version '{version_tag}'")
+
+
+def load_model(version_tag: str):
+    """
+    Load a saved model by version tag.
+
+    Args:
+        version_tag: e.g. 'v1_base', 'v2_tuned'
+
+    Returns:
+        Loaded sklearn model
+    """
+    metadata = _load_metadata()
+
+    if version_tag not in metadata:
+        raise KeyError(f"[model_utils] Version '{version_tag}' not found in metadata registry.")
+
+    filename = metadata[version_tag]["filename"]
+    path = os.path.join(MODEL_DIR, filename)
+
+    if not os.path.exists(path):
+        raise FileNotFoundError(f"[model_utils] Model file not found at {path}")
+
+    model = joblib.load(path)
+    print(f"[model_utils] Loaded model version '{version_tag}' from {path}")
+    return model
+
+
+def list_versions():
+    """
+    Print all saved model versions with their metadata.
+    """
+    metadata = _load_metadata()
+
+    if not metadata:
+        print("[model_utils] No saved model versions found.")
+        return
+
+    print(f"\n{'='*50}")
+    print(f"{'VERSION':<15} {'DATE':<22} {'ACCURACY':<10}")
+    print(f"{'='*50}")
+    for version, info in metadata.items():
+        accuracy = info["metrics"].get("accuracy", "N/A")
+        print(f"{version:<15} {info['training_date']:<22} {accuracy:<10}")
+    print(f"{'='*50}\n")
+
+
+def rollback(version_tag: str):
+    """
+    Rollback to a specific model version by loading it.
+
+    Args:
+        version_tag: Version to roll back to
+
+    Returns:
+        Loaded sklearn model
+    """
+    print(f"[model_utils] Rolling back to version '{version_tag}'...")
+    return load_model(version_tag)