diff --git a/examples/neptunelogger.py b/examples/neptunelogger.py
new file mode 100644
index 000000000..bbe82f725
--- /dev/null
+++ b/examples/neptunelogger.py
@@ -0,0 +1,57 @@
+"""Example application with a dummy model using callback-based Neptune integration.
+"""
+import numpy as np
+
+from keras_contrib import callbacks
+from keras.models import Sequential
+from keras.layers import Dense
+
+
+# Replace below with your own credentials
+# If NEPTUNE_API_TOKEN is None, read from environment variable
+#  otherwise you should provide it as a str
+PROJECT_QUALIFIED_NAME = "your_username/your_project"
+EXPERIMENT_NAME = "test-neptunelogger"
+NEPTUNE_API_TOKEN = None
+
+
+def build_model():
+    """Build a dummy binary classification model model.
+
+    Returns:
+        Keras.models.Model: The dummy model.
+    """
+    model = Sequential([
+        Dense(2, activation='relu', input_shape=(2,)),
+        Dense(1, activation='sigmoid')
+    ])
+    return model
+
+
+def test_NeptuneLogger():
+    """Test the NeptuneLogger callback with a dummy model and dataset.
+    """
+    X = np.random.rand(100, 2)
+    y = np.random.rand(100).reshape(-1, 1)
+
+    model = build_model()
+    model.compile(
+        optimizer='sgd',
+        loss='binary_crossentropy',
+        metrics=['accuracy']
+    )
+    model.fit(
+        X, y,
+        batch_size=1,
+        epochs=1,
+        verbose=0,
+        callbacks=[
+            callbacks.NeptuneLogger(
+                project_qualified_name=PROJECT_QUALIFIED_NAME,
+                experiment_name=EXPERIMENT_NAME,
+                api_token=NEPTUNE_API_TOKEN
+            )
+        ])
+
+if __name__ == '__main__':
+    test_NeptuneLogger()
diff --git a/keras_contrib/callbacks/__init__.py b/keras_contrib/callbacks/__init__.py
index 96d82c19e..8a99b9bc0 100644
--- a/keras_contrib/callbacks/__init__.py
+++ b/keras_contrib/callbacks/__init__.py
@@ -2,3 +2,4 @@
 from .dead_relu_detector import DeadReluDetector
 from .cyclical_learning_rate import CyclicLR
 from .tensorboard import TensorBoardGrouped
+from .neptune import NeptuneLogger
diff --git a/keras_contrib/callbacks/neptune.py b/keras_contrib/callbacks/neptune.py
new file mode 100644
index 000000000..5545052a5
--- /dev/null
+++ b/keras_contrib/callbacks/neptune.py
@@ -0,0 +1,64 @@
+"""Neptune integration for Keras via callback.
+"""
+from __future__ import absolute_import
+from __future__ import print_function
+
+from keras.callbacks import Callback
+
+import neptune
+
+
+class NeptuneLogger(Callback):
+    """
+    Neptune integration with keras callbacks.
+    """
+
+    def __init__(
+            self,
+            project_qualified_name, experiment_name, api_token=None, **kwargs):
+        """
+        Construct the NeptuneLogger callback and log in to neptune.
+        Neptune experiment creation is delayed until the beginning of training.
+
+        Args:
+            project_qualified_name (str): The username and project name with
+                which to log in.
+            experiment_name (str): The name to give to the new experiment.
+            api_token (str, optional): The Neptune API token to use as
+                credentials. By default this is read from the environment
+                variable NEPTUNE_API_TOKEN.
+            **kwargs:
+                For full list see neptune.projects.Project.create_experiment.
+                Some useful keyword names include: description (str),
+                params (dict), properties (dict), tags (list of str),
+                upload_source_files (list of str paths).
+                Note that the value of this can change between instatiation
+                and on_train_begin.
+
+        """
+        self.experiment_name = experiment_name
+        self.experiment_kwargs = kwargs
+        self.project_qualified_name = project_qualified_name
+        self.experiment = None
+
+        self.session = neptune.sessions.Session(api_token=api_token)
+        self.project = self.session.get_project(project_qualified_name)
+
+    def __del__(self):
+        if self.experiment:
+            self.experiment.stop()
+
+    def on_train_begin(self, logs):
+        if self.experiment:
+            return
+        self.experiment = self.project.create_experiment(
+            self.experiment_name, **self.experiment_kwargs)
+
+    def on_epoch_end(self, epoch, logs=None):
+        if not logs:
+            return
+        for key, value in logs.items():
+            try:
+                self.experiment.send_metric(key, epoch, float(value))
+            except ValueError:
+                pass  # Ignore non numeric values