diff --git a/cloudbuild/python-scripts/insert_dynamic_partitions.py b/cloudbuild/python-scripts/insert_dynamic_partitions.py
new file mode 100644
index 00000000..d1de8a42
--- /dev/null
+++ b/cloudbuild/python-scripts/insert_dynamic_partitions.py
@@ -0,0 +1,144 @@
+"""Python script to dynamically partitions to a BigQuery partitioned table."""
+
+import argparse
+from collections.abc import Sequence
+import datetime
+import logging
+import threading
+import time
+from absl import app
+from utils import utils
+
+
+def sleep_for_seconds(duration):
+    logging.info(
+        'Going to sleep, waiting for connector to read existing, Time: %s',
+        datetime.datetime.now()
+    )
+    # Buffer time to ensure that new partitions are created
+    # after previous read session and before next split discovery.
+    time.sleep(duration)
+
+
+def main(argv: Sequence[str]) -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        '--refresh_interval',
+        dest='refresh_interval',
+        help='Minutes between checking new data',
+        type=int,
+        required=True,
+    )
+    parser.add_argument(
+        '--project_name',
+        dest='project_name',
+        help='Project Id which contains the table to be read.',
+        type=str,
+        required=True,
+    )
+    parser.add_argument(
+        '--dataset_name',
+        dest='dataset_name',
+        help='Dataset Name which contains the table to be read.',
+        type=str,
+        required=True,
+    )
+    parser.add_argument(
+        '--table_name',
+        dest='table_name',
+        help='Table Name of the table which is read in the test.',
+        type=str,
+        required=True,
+    )
+    parser.add_argument(
+        '-n',
+        '--number_of_rows_per_partition',
+        dest='number_of_rows_per_partition',
+        help='Number of rows to insert per partition.',
+        type=int,
+        required=False,
+        default=30000,
+    )
+
+    args = parser.parse_args(argv[1:])
+
+    # Providing the values.
+    project_name = args.project_name
+    dataset_name = args.dataset_name
+    table_name = args.table_name
+    number_of_rows_per_partition = args.number_of_rows_per_partition
+
+    execution_timestamp = datetime.datetime.now(tz=datetime.timezone.utc).replace(
+        hour=0, minute=0, second=0, microsecond=0
+    )
+    refresh_interval = int(args.refresh_interval)
+
+    # Set the partitioned table.
+    table_id = f'{project_name}.{dataset_name}.{table_name}'
+
+    # Now add the partitions to the table.
+    # Hardcoded schema. Needs to be same as that in the pre-created table.
+    simple_avro_schema_fields_string = (
+        '"fields": [{"name": "name", "type": "string"},{"name": "number",'
+        '"type": "long"},{"name" : "ts", "type" : {"type" :'
+        '"long","logicalType": "timestamp-micros"}}]'
+    )
+    simple_avro_schema_string = (
+        '{"namespace": "project.dataset","type": "record","name":'
+        ' "table","doc": "Avro Schema for project.dataset.table",'
+        f'{simple_avro_schema_fields_string}'
+        '}'
+    )
+
+    # hardcoded for e2e test.
+    # partitions[i] * number_of_rows_per_partition are inserted per phase.
+    partitions = [2, 1, 2]
+    #  BQ rate limit is exceeded due to large number of rows.
+    number_of_threads = 2
+    number_of_rows_per_thread = number_of_rows_per_partition // number_of_threads
+
+    avro_file_local = 'mockData.avro'
+    table_creation_utils = utils.TableCreationUtils(
+        simple_avro_schema_string,
+        number_of_rows_per_thread,
+        table_id,
+    )
+
+    # Insert iteratively.
+    prev_partitions_offset = 0
+    for number_of_partitions in partitions:
+        start_time = time.time()
+        # Wait for read stream formation.
+        sleep_for_seconds(2.5 * 60)
+
+        # This represents one iteration.
+        for partition_number in range(number_of_partitions):
+            threads = list()
+            # Insert via concurrent threads.
+            for thread_number in range(number_of_threads):
+                avro_file_local_identifier = avro_file_local.replace(
+                    '.', '_' + str(thread_number) + '.'
+                )
+                thread = threading.Thread(
+                    target=table_creation_utils.avro_to_bq_with_cleanup,
+                    kwargs={
+                        'avro_file_local_identifier': avro_file_local_identifier,
+                        'partition_number': partition_number + prev_partitions_offset,
+                        'current_timestamp': execution_timestamp,
+                    },
+                )
+                threads.append(thread)
+                thread.start()
+            for _, thread in enumerate(threads):
+                thread.join()
+
+        time_elapsed = time.time() - start_time
+        prev_partitions_offset += number_of_partitions
+
+        # We wait until the read streams are formed again.
+        # So that the records just created can be read.
+        sleep_for_seconds(float(60 * refresh_interval) - time_elapsed)
+
+
+if __name__ == '__main__':
+    app.run(main)