diff --git a/runtime/onert/api/python/package/experimental/train/optimizer/__init__.py b/runtime/onert/api/python/package/experimental/train/optimizer/__init__.py
new file mode 100644
index 00000000000..ae450a7d0f4
--- /dev/null
+++ b/runtime/onert/api/python/package/experimental/train/optimizer/__init__.py
@@ -0,0 +1,5 @@
+from .sgd import SGD
+from .adam import Adam
+from onert.native.libnnfw_api_pybind import trainable_ops
+
+__all__ = ["SGD", "Adam", "trainable_ops"]
diff --git a/runtime/onert/api/python/package/experimental/train/optimizer/adam.py b/runtime/onert/api/python/package/experimental/train/optimizer/adam.py
new file mode 100644
index 00000000000..29de7d75517
--- /dev/null
+++ b/runtime/onert/api/python/package/experimental/train/optimizer/adam.py
@@ -0,0 +1,20 @@
+from .optimizer import Optimizer
+
+
+class Adam(Optimizer):
+    """
+    Adam optimizer.
+    """
+    def __init__(self, learning_rate=0.001, beta1=0.9, beta2=0.999, epsilon=1e-7):
+        """
+        Initialize the Adam optimizer.
+        Args:
+            learning_rate (float): The learning rate for optimization.
+            beta1 (float): Exponential decay rate for the first moment estimates.
+            beta2 (float): Exponential decay rate for the second moment estimates.
+            epsilon (float): Small constant to prevent division by zero.
+        """
+        super().__init__(learning_rate)
+        self.beta1 = beta1
+        self.beta2 = beta2
+        self.epsilon = epsilon
diff --git a/runtime/onert/api/python/package/experimental/train/optimizer/optimizer.py b/runtime/onert/api/python/package/experimental/train/optimizer/optimizer.py
new file mode 100644
index 00000000000..5f47ddaf4bd
--- /dev/null
+++ b/runtime/onert/api/python/package/experimental/train/optimizer/optimizer.py
@@ -0,0 +1,15 @@
+from onert.native.libnnfw_api_pybind import trainable_ops
+
+
+class Optimizer:
+    """
+    Base class for optimizers. Subclasses should implement the `step` method.
+    """
+    def __init__(self, learning_rate=0.001, nums_trainable_ops=trainable_ops.ALL):
+        """
+        Initialize the optimizer.
+        Args:
+            learning_rate (float): The learning rate for optimization.
+        """
+        self.learning_rate = learning_rate
+        self.nums_trainable_ops = nums_trainable_ops
diff --git a/runtime/onert/api/python/package/experimental/train/optimizer/sgd.py b/runtime/onert/api/python/package/experimental/train/optimizer/sgd.py
new file mode 100644
index 00000000000..6e08f927573
--- /dev/null
+++ b/runtime/onert/api/python/package/experimental/train/optimizer/sgd.py
@@ -0,0 +1,21 @@
+from .optimizer import Optimizer
+
+
+class SGD(Optimizer):
+    """
+    Stochastic Gradient Descent (SGD) optimizer.
+    """
+    def __init__(self, learning_rate=0.001, momentum=0.0):
+        """
+        Initialize the SGD optimizer.
+        Args:
+            learning_rate (float): The learning rate for optimization.
+            momentum (float): Momentum factor (default: 0.0).
+        """
+        super().__init__(learning_rate)
+
+        if momentum != 0.0:
+            raise NotImplementedError(
+                "Momentum is not supported in the current version of SGD.")
+        self.momentum = momentum
+        self.velocity = None