diff --git a/optimization/adam.py b/optimization/adam.py
new file mode 100644
index 0000000..8b824d9
--- /dev/null
+++ b/optimization/adam.py
@@ -0,0 +1,68 @@
+import tensorflow as tf
+from typing import Callable
+class Adam:
+    def __init__(self,
+                 learning_rate:float,
+                 beta1:float,
+                 beta2:float,
+                 objective_function:Callable,
+                 params:tf.Tensor,
+                 max_iteration:int) -> None:
+    
+        self.learning_rate = learning_rate
+        self.beta1 = beta1
+        self.beta2 = beta2
+        self.objective_fn =objective_function
+        self.params = params
+        self.max_iteration = max_iteration
+        self.first_momentum = tf.zeros_like(self.params)
+        self.second_momentum = tf.zeros_like(self.params)
+        self.epsilon = 1e-6
+        self.losses = []
+
+    def optimize(self):
+        for t in range(1,self.max_iteration+1):
+
+            with tf.GradientTape() as tape:
+                loss = self.objective_fn(self.params)
+                self.losses.append(loss.numpy())
+            grad = tape.gradient(loss,self.params)
+            grad_squared = tf.square(grad)
+            
+            first_momentum_update = self.first_momentum*self.beta1 + grad * (1-self.beta1)
+            second_momentum_update = self.second_momentum * self.beta2 + grad_squared* (1-self.beta2)
+            
+            self.first_momentum= first_momentum_update
+            self.second_momentum = second_momentum_update
+
+            first_momentum_cap = self.first_momentum/(1-tf.pow(self.beta1,t))
+            second_momentum_cap = self.second_momentum/(1-tf.pow(self.beta2,t))
+            
+            step = first_momentum_cap*self.learning_rate/tf.sqrt(second_momentum_cap+self.epsilon)
+
+            self.params.assign_sub(step)
+
+def quadratic_loss(params):
+    return tf.reduce_sum(tf.square(params))
+if __name__ == '__main__':
+
+    # Define initial parameters
+    initial_params = tf.Variable([1.0, 2.0, 3.0], dtype=tf.float32)
+
+    # Adam optimizer parameters
+    learning_rate = 0.01
+    beta1 = 0.9
+    beta2 = 0.999
+    max_iteration = 1000
+
+    # Create an instance of Adam optimizer
+    optimizer = Adam(learning_rate, beta1, beta2, quadratic_loss, initial_params, max_iteration)
+
+    # Optimize the parameters
+    optimizer.optimize()
+
+    # Get the optimized parameters
+    optimized_params = optimizer.params.numpy()
+
+    print(optimized_params)
+    print(optimizer.losses)
\ No newline at end of file