diff --git a/platform/micro-services/command-line-interface/source/cli/peoples-speech.py b/platform/micro-services/command-line-interface/source/cli/peoples-speech.py
index ab16d058..f67d34db 100644
--- a/platform/micro-services/command-line-interface/source/cli/peoples-speech.py
+++ b/platform/micro-services/command-line-interface/source/cli/peoples-speech.py
@@ -3,26 +3,42 @@
 
 from argparse import ArgumentParser
 
-import peoples_speech.data_book
-import peoples_speech.data_export
-import peoples_speech.task_manager
-
 import config
 
 import logging
 
+import peoples_speech.data_book
+import peoples_speech.data_export
+import peoples_speech.task_manager
+
 logger = logging.getLogger(__name__)
 
 def main():
     parser = ArgumentParser("The MLCommons data engineering framework.")
 
+    subparsers = parser.add_subparsers()
+
+    setup_cli_parser(subparsers)
+    peoples_speech.data_export.setup_cli_parser(subparsers)
+
+    args = parser.parse_args()
+
+    args.func(args)
+
+def setup_cli_parser(subparsers):
+
+    parser = subparsers.add_parser('dataset')
+
     parser.add_argument("-i", "--data-book-path", default=sample_databook_path(), help="Path to data book to generate a dataset for.")
     parser.add_argument("-o", "--output-dataset-path", default="", help="The path to save the new dataset.")
-    parser.add_argument("-c", "--config-file-path", default=".csv", help="The path to save the new dataset.")
+    parser.add_argument("-c", "--config-file-path", default=".csv", help="The path to the config file.")
     parser.add_argument("-v", "--verbose", default=False, action="store_true", help="Print out debug messages.")
     parser.add_argument("-vi", "--verbose-info", default=False, action="store_true", help="Print out info messages.")
 
-    arguments = vars(parser.parse_args())
+    parser.set_defaults(func=dispatch)
+
+def dispatch(args):
+    arguments = vars(args)
 
     config = setup_config(arguments)
 
diff --git a/platform/micro-services/data-export/build-scripts/flask/cloud-build/cloudbuild.sh b/platform/micro-services/data-export/build-scripts/flask/cloud-build/cloudbuild.sh
new file mode 100755
index 00000000..90fb70f4
--- /dev/null
+++ b/platform/micro-services/data-export/build-scripts/flask/cloud-build/cloudbuild.sh
@@ -0,0 +1,13 @@
+#! /bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# u unset variables are errors
+# f disable globbing on *
+# pipefail | produces a failure code if any stage fails
+set -euf -o pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+gcloud builds submit --config cloudbuild.yaml $LOCAL_DIRECTORY/../../../../../..
diff --git a/platform/micro-services/data-export/build-scripts/flask/cloud-build/cloudbuild.yaml b/platform/micro-services/data-export/build-scripts/flask/cloud-build/cloudbuild.yaml
new file mode 100644
index 00000000..662f159e
--- /dev/null
+++ b/platform/micro-services/data-export/build-scripts/flask/cloud-build/cloudbuild.yaml
@@ -0,0 +1,14 @@
+steps:
+# build the data export container
+- name: 'gcr.io/cloud-builders/docker'
+  args: [ 'build', '-t', 'gcr.io/the-peoples-speech/data-export:0.12', '-f', 'platform/micro-services/data-export/build-scripts/flask/docker/Dockerfile', 'platform/micro-services/data-export' ]
+# push container image
+- name: "gcr.io/cloud-builders/docker"
+  args: ["push", "gcr.io/the-peoples-speech/data-export:0.12"]
+# deploy container image to GKE staging
+- name: "gcr.io/cloud-builders/gke-deploy"
+  args:
+  - run
+  - --filename=platform/micro-services/data-export/build-scripts/flask/kubernetes/export-service.yaml
+  - --location=europe-west4-c
+  - --cluster=peoples-speech-platform
diff --git a/platform/micro-services/data-export/build-scripts/flask/docker/Dockerfile b/platform/micro-services/data-export/build-scripts/flask/docker/Dockerfile
new file mode 100644
index 00000000..b8be3999
--- /dev/null
+++ b/platform/micro-services/data-export/build-scripts/flask/docker/Dockerfile
@@ -0,0 +1,9 @@
+FROM python
+
+COPY . /app
+
+ENV BOTO_CONFIG=/app/source/configs/google.boto
+
+EXPOSE 5000
+
+CMD /app/start-production $PORT
diff --git a/platform/micro-services/data-export/build-scripts/flask/docker/build-container.sh b/platform/micro-services/data-export/build-scripts/flask/docker/build-container.sh
new file mode 100755
index 00000000..b75fa3e6
--- /dev/null
+++ b/platform/micro-services/data-export/build-scripts/flask/docker/build-container.sh
@@ -0,0 +1,14 @@
+#! /bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# u unset variables are errors
+# f disable globbing on *
+# pipefail | produces a failure code if any stage fails
+set -euf -o pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+docker build -t data-export:0.1 -f $LOCAL_DIRECTORY/Dockerfile $LOCAL_DIRECTORY/../../../../../..
+
diff --git a/platform/micro-services/data-export/build-scripts/flask/kubernetes/deploy-staging.sh b/platform/micro-services/data-export/build-scripts/flask/kubernetes/deploy-staging.sh
new file mode 100644
index 00000000..78874bb2
--- /dev/null
+++ b/platform/micro-services/data-export/build-scripts/flask/kubernetes/deploy-staging.sh
@@ -0,0 +1,2 @@
+gcloud container clusters get-credentials peoples-speech-platform
+kubectl create deployment data-export --image=gcr.io/peoples-speech/data-export:latest
diff --git a/platform/micro-services/data-export/build-scripts/flask/kubernetes/export-service.yaml b/platform/micro-services/data-export/build-scripts/flask/kubernetes/export-service.yaml
new file mode 100644
index 00000000..05b5fb7c
--- /dev/null
+++ b/platform/micro-services/data-export/build-scripts/flask/kubernetes/export-service.yaml
@@ -0,0 +1,50 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  labels:
+    app: data-export
+  name: data-export
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: data-export
+  template:
+    metadata:
+      labels:
+        app: data-export
+    spec:
+      containers:
+      - image: gcr.io/the-peoples-speech/data-export:0.12
+        name: data-export
+        ports:
+        - containerPort: 5000
+          name: tcp-c-5000
+        env:
+        - name: "PORT"
+          value: "5000"
+        volumeMounts:
+          - mountPath: "/app/credentials"
+            name: gcloud-service-account-key
+            readOnly: true
+      volumes:
+        - name: gcloud-service-account-key
+          secret:
+            secretName: gcloud-service-account-key
+---
+apiVersion: v1
+kind: Service
+metadata:
+  labels:
+    app: data-export
+  name: data-export
+spec:
+  selector:
+    app: data-export
+  ports:
+    - port: 5000
+      targetPort: 5000
+      protocol: TCP
+      name: tcp-5000
+  type: LoadBalancer
+
diff --git a/platform/micro-services/data-export/build-scripts/flask/kubernetes/run-export-service.yaml b/platform/micro-services/data-export/build-scripts/flask/kubernetes/run-export-service.yaml
new file mode 100644
index 00000000..3c0d7a46
--- /dev/null
+++ b/platform/micro-services/data-export/build-scripts/flask/kubernetes/run-export-service.yaml
@@ -0,0 +1,21 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  labels:
+    app.kubernetes.io/name: data-export
+  name: data-export
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: data-export
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: data-export
+    spec:
+      containers:
+      - image: gcr.io/peoples-speech/data-export
+        name: data-export
+        ports:
+        - containerPort: 8080
diff --git a/platform/micro-services/data-export/requirements.txt b/platform/micro-services/data-export/requirements.txt
new file mode 100644
index 00000000..2e86cc7c
--- /dev/null
+++ b/platform/micro-services/data-export/requirements.txt
@@ -0,0 +1,7 @@
+flask
+flask_cors
+python-configuration[yaml]
+smart_open[gcs]
+redis
+gsutil
+google-compute-engine
diff --git a/platform/micro-services/data-export/run-tests b/platform/micro-services/data-export/run-tests
new file mode 100755
index 00000000..82f0e4a1
--- /dev/null
+++ b/platform/micro-services/data-export/run-tests
@@ -0,0 +1,41 @@
+#! /bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# u unset variables are errors
+# f disable globbing on *
+# pipefail | produces a failure code if any stage fails
+set -euf -o pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+# Setup virtual environment
+PYTHON_ENV=$(python3 -c "import sys; sys.stdout.write(sys.prefix) if (hasattr(sys, 'real_prefix') or sys.base_prefix != sys.prefix) else sys.stdout.write('0')")
+if [[ $PYTHON_ENV == 0 ]];
+then
+echo "Not in virtual environment"
+
+ACTIVATE=$LOCAL_DIRECTORY/environment/bin/activate
+
+if [ ! -f $ACTIVATE ]; then
+echo "Virtual environment doesn't exist, making it..."
+python3 -m venv $LOCAL_DIRECTORY/environment
+python3 -m pip install --upgrade pip > /dev/null
+fi
+
+source $ACTIVATE
+else
+echo "Running in virtual environment $PYTHON_ENV"
+fi
+
+# Make sure requirements are installed
+pip install -r $LOCAL_DIRECTORY/requirements.txt > /dev/null
+
+# Set python environment
+PYTHONPATH+="$LOCAL_DIRECTORY/source"
+export PYTHONPATH
+
+python source/data_export/api/test/test_export_dataset.py
+
+
diff --git a/platform/micro-services/data-export/source/add-clean-text.py b/platform/micro-services/data-export/source/commands/add-clean-text.py
similarity index 100%
rename from platform/micro-services/data-export/source/add-clean-text.py
rename to platform/micro-services/data-export/source/commands/add-clean-text.py
diff --git a/platform/micro-services/data-export/source/add-duration-metadata.py b/platform/micro-services/data-export/source/commands/add-duration-metadata.py
similarity index 100%
rename from platform/micro-services/data-export/source/add-duration-metadata.py
rename to platform/micro-services/data-export/source/commands/add-duration-metadata.py
diff --git a/platform/micro-services/data-export/source/add-google-speech-transcript.py b/platform/micro-services/data-export/source/commands/add-google-speech-transcript.py
similarity index 100%
rename from platform/micro-services/data-export/source/add-google-speech-transcript.py
rename to platform/micro-services/data-export/source/commands/add-google-speech-transcript.py
diff --git a/platform/micro-services/data-export/source/add-librispeech-metadata.py b/platform/micro-services/data-export/source/commands/add-librispeech-metadata.py
similarity index 100%
rename from platform/micro-services/data-export/source/add-librispeech-metadata.py
rename to platform/micro-services/data-export/source/commands/add-librispeech-metadata.py
diff --git a/platform/micro-services/data-export/source/convert-audio-format.py b/platform/micro-services/data-export/source/commands/convert-audio-format.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-audio-format.py
rename to platform/micro-services/data-export/source/commands/convert-audio-format.py
diff --git a/platform/micro-services/data-export/source/convert-cc-search.py b/platform/micro-services/data-export/source/commands/convert-cc-search.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-cc-search.py
rename to platform/micro-services/data-export/source/commands/convert-cc-search.py
diff --git a/platform/micro-services/data-export/source/convert-common-voice.py b/platform/micro-services/data-export/source/commands/convert-common-voice.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-common-voice.py
rename to platform/micro-services/data-export/source/commands/convert-common-voice.py
diff --git a/platform/micro-services/data-export/source/convert-dipco-dataset-format.py b/platform/micro-services/data-export/source/commands/convert-dipco-dataset-format.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-dipco-dataset-format.py
rename to platform/micro-services/data-export/source/commands/convert-dipco-dataset-format.py
diff --git a/platform/micro-services/data-export/source/convert-librispeech.py b/platform/micro-services/data-export/source/commands/convert-librispeech.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-librispeech.py
rename to platform/micro-services/data-export/source/commands/convert-librispeech.py
diff --git a/platform/micro-services/data-export/source/convert-librivox-google-cloud.py b/platform/micro-services/data-export/source/commands/convert-librivox-google-cloud.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-librivox-google-cloud.py
rename to platform/micro-services/data-export/source/commands/convert-librivox-google-cloud.py
diff --git a/platform/micro-services/data-export/source/convert-librivox.py b/platform/micro-services/data-export/source/commands/convert-librivox.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-librivox.py
rename to platform/micro-services/data-export/source/commands/convert-librivox.py
diff --git a/platform/micro-services/data-export/source/convert-timit-dataset-format.py b/platform/micro-services/data-export/source/commands/convert-timit-dataset-format.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-timit-dataset-format.py
rename to platform/micro-services/data-export/source/commands/convert-timit-dataset-format.py
diff --git a/platform/micro-services/data-export/source/convert-voicery.py b/platform/micro-services/data-export/source/commands/convert-voicery.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-voicery.py
rename to platform/micro-services/data-export/source/commands/convert-voicery.py
diff --git a/platform/micro-services/data-export/source/convert-warc-to-csv.py b/platform/micro-services/data-export/source/commands/convert-warc-to-csv.py
similarity index 100%
rename from platform/micro-services/data-export/source/convert-warc-to-csv.py
rename to platform/micro-services/data-export/source/commands/convert-warc-to-csv.py
diff --git a/platform/micro-services/data-export/source/count-duration.py b/platform/micro-services/data-export/source/commands/count-duration.py
similarity index 100%
rename from platform/micro-services/data-export/source/count-duration.py
rename to platform/micro-services/data-export/source/commands/count-duration.py
diff --git a/platform/micro-services/data-export/source/make-librivox-train-test-split.py b/platform/micro-services/data-export/source/commands/make-librivox-train-test-split.py
similarity index 100%
rename from platform/micro-services/data-export/source/make-librivox-train-test-split.py
rename to platform/micro-services/data-export/source/commands/make-librivox-train-test-split.py
diff --git a/platform/micro-services/data-export/source/make-peoples-speech-train-test-splits.py b/platform/micro-services/data-export/source/commands/make-peoples-speech-train-test-splits.py
similarity index 100%
rename from platform/micro-services/data-export/source/make-peoples-speech-train-test-splits.py
rename to platform/micro-services/data-export/source/commands/make-peoples-speech-train-test-splits.py
diff --git a/platform/micro-services/data-export/source/make-tar-google-cloud.py b/platform/micro-services/data-export/source/commands/make-tar-google-cloud.py
similarity index 100%
rename from platform/micro-services/data-export/source/make-tar-google-cloud.py
rename to platform/micro-services/data-export/source/commands/make-tar-google-cloud.py
diff --git a/platform/micro-services/data-export/source/configs/default.yaml b/platform/micro-services/data-export/source/configs/default.yaml
new file mode 100644
index 00000000..2d3b18ac
--- /dev/null
+++ b/platform/micro-services/data-export/source/configs/default.yaml
@@ -0,0 +1,6 @@
+exporter:
+    type: CloudExporter
+    endpoint: http://34.91.68.228
+
+verbose: True
+
diff --git a/platform/micro-services/data-export/source/configs/local.yaml b/platform/micro-services/data-export/source/configs/local.yaml
new file mode 100644
index 00000000..e5278a66
--- /dev/null
+++ b/platform/micro-services/data-export/source/configs/local.yaml
@@ -0,0 +1,10 @@
+
+datasets:
+    "0": gs://the-peoples-speech-west-europe/peoples-speech-v0.8/unittest.csv
+
+exporter:
+    type: LocalExporter
+
+output_dataset_path: gs://the-peoples-speech-west-europe/peoples-speech-v0.8/unittest.tar.gz
+
+verbose: True
diff --git a/platform/micro-services/data-export/source/configs/server.yaml b/platform/micro-services/data-export/source/configs/server.yaml
new file mode 100644
index 00000000..956e69b2
--- /dev/null
+++ b/platform/micro-services/data-export/source/configs/server.yaml
@@ -0,0 +1,10 @@
+exporter:
+    type: GoogleCloudParallelExporter
+    task_count: 1024
+
+google:
+    work_queue:
+        name: "export_work_queue"
+
+verbose: True
+
diff --git a/platform/micro-services/data-export/source/data_export/__init__.py b/platform/micro-services/data-export/source/data_export/__init__.py
index b0596091..9ca01754 100644
--- a/platform/micro-services/data-export/source/data_export/__init__.py
+++ b/platform/micro-services/data-export/source/data_export/__init__.py
@@ -1,5 +1,7 @@
 
 from data_export.api.save_dataset import save_dataset
-
+from data_export.api.export_dataset_by_id import export_dataset_by_id
+from data_export.api.export_dataset import export_dataset
+from data_export.api.setup_cli_parser import setup_cli_parser
 
 
diff --git a/platform/micro-services/data-export/source/data_export/api/export_dataset.py b/platform/micro-services/data-export/source/data_export/api/export_dataset.py
new file mode 100644
index 00000000..387daa3c
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/api/export_dataset.py
@@ -0,0 +1,13 @@
+
+from data_export.export import ExporterFactory
+from data_export.utility.get_config import get_config
+
+<<<<<<< HEAD
+def export_dataset(output_dataset, dataset, config = get_config()):
+    exporter = ExporterFactory(config).create()
+    return exporter.export(output_dataset, dataset)
+=======
+def export_dataset(dataset, config = get_config()):
+    exporter = ExporterFactory(config).create()
+    exporter.export(dataset)
+>>>>>>> d17bf137... TEST: Added unit test for data export.
diff --git a/platform/micro-services/data-export/source/data_export/api/export_dataset_by_id.py b/platform/micro-services/data-export/source/data_export/api/export_dataset_by_id.py
new file mode 100644
index 00000000..8605488f
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/api/export_dataset_by_id.py
@@ -0,0 +1,10 @@
+
+from data_export.export import ExporterFactory
+from data_export.database import get_dataset_from_id
+from data_export.utility.get_config import get_config
+
+def export_dataset_by_id(dataset_id, config = get_config()):
+    exporter = ExporterFactory(config).create()
+    dataset = get_dataset_from_id(config, dataset_id)
+    exporter.export(dataset)
+
diff --git a/platform/micro-services/data-export/source/data_export/api/setup_cli_parser.py b/platform/micro-services/data-export/source/data_export/api/setup_cli_parser.py
new file mode 100644
index 00000000..d5a9023f
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/api/setup_cli_parser.py
@@ -0,0 +1,70 @@
+from data_export.api.export_dataset import export_dataset
+
+import config
+import logging
+import os
+
+logger = logging.getLogger(__name__)
+
+
+def setup_cli_parser(subparsers):
+
+    parser = subparsers.add_parser('export')
+
+    parser.add_argument("-i", "--input-dataset", default="gs://the-peoples-speech-west-europe/peoples-speech-v0.8/unittest.csv", help="The dataset to export.")
+    parser.add_argument("-o", "--output-dataset-path", default="gs://the-peoples-speech-west-europe/peoples-speech-v0.8/unittest.tar.gz", help="The path to save the new dataset.")
+    parser.add_argument("-c", "--config-file-path", default="", help="The path to the config file.")
+    parser.add_argument("-v", "--verbose", default=False, action="store_true", help="Print out debug messages.")
+    parser.add_argument("-vi", "--verbose-info", default=False, action="store_true", help="Print out info messages.")
+
+    parser.set_defaults(func=dispatch)
+
+def dispatch(args):
+    arguments = vars(args)
+
+    config = setup_config(arguments)
+
+    setup_logging(config)
+
+    logger.debug("Full config: " + str(config))
+
+    export_dataset(config["output_dataset_path"], config["input_dataset"], config)
+
+def setup_config(dictionary):
+    return config.ConfigurationSet(
+        config.config_from_env(prefix="MLCOMMONS"),
+        config.config_from_yaml(config_path(dictionary), read_from_file=True),
+        config.config_from_dict(dictionary),
+    )
+
+def config_path(dictionary):
+    if os.path.exists(dictionary["config_file_path"]):
+        return dictionary["config_file_path"]
+
+    home = os.path.expanduser("~")
+    home_config_path = os.path.join(home, ".mlcommons", "config.yaml")
+    if os.path.exists(home_config_path):
+        return home_config_path
+
+    return os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(__file__))), "configs", "default.yaml")
+
+def setup_logging(arguments):
+
+    logging_format = "%(asctime)s - %(levelname)s - %(name)s - %(message)s"
+
+    if arguments["verbose"]:
+        logging.basicConfig(level=logging.DEBUG, format=logging_format)
+    elif arguments["verbose_info"]:
+        logging.basicConfig(level=logging.INFO, format=logging_format)
+    else:
+        logging.basicConfig(level=logging.WARNING, format=logging_format)
+
+    root_logger = logging.getLogger()
+
+    if arguments["verbose"]:
+        root_logger.setLevel(logging.DEBUG)
+    elif arguments["verbose_info"]:
+        root_logger.setLevel(logging.INFO)
+    else:
+        root_logger.setLevel(logging.WARNING)
+
diff --git a/platform/micro-services/data-export/source/data_export/api/test/test_export_dataset.py b/platform/micro-services/data-export/source/data_export/api/test/test_export_dataset.py
new file mode 100644
index 00000000..932a0c21
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/api/test/test_export_dataset.py
@@ -0,0 +1,36 @@
+import unittest
+import tarfile
+import io
+from data_export.api.export_dataset import export_dataset
+from data_export.utility.load_dataset_csv import load_dataset_csv
+from data_export.utility.load_dataset_csv import load_dataset_csv_from_file
+
+class TestDataExport(unittest.TestCase):
+
+    def test_tiny_local(self):
+        tiny_dataset = "gs://the-peoples-speech-west-europe/peoples-speech-v0.8/unittest.csv"
+        tiny_config = {
+            "exporter" : { "type" : "LocalExporter" } }
+
+        export_dataset("/tmp/export-data/unittest.tar.gz", tiny_dataset, tiny_config)
+
+        self.extract_dataset(tiny_config)
+        self.verify_dataset(tiny_dataset, tiny_config)
+
+    def extract_dataset(self, config):
+        with tarfile.open(config["output_dataset_path"], "r:gz") as tar:
+            members = tar.getmembers()
+
+    def verify_dataset(self, dataset, config):
+
+        samples = load_dataset_csv(dataset, task_id=0, task_count=1)
+        with tarfile.open(config["output_dataset_path"], "r:gz") as tar:
+            samples_file = io.TextIOWrapper(tar.extractfile("export-data/dataset.csv"))
+            loaded_samples = load_dataset_csv_from_file(samples_file, task_id=0, task_count=1)
+
+            for original_sample, new_sample in zip(samples, loaded_samples):
+                self.assertEqual(original_sample, new_sample)
+
+
+if __name__ == '__main__':
+    unittest.main()
diff --git a/platform/micro-services/data-export/source/data_export/database/__init__.py b/platform/micro-services/data-export/source/data_export/database/__init__.py
new file mode 100644
index 00000000..cfeed4f8
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/database/__init__.py
@@ -0,0 +1,4 @@
+
+from data_export.database.get_dataset_from_id import get_dataset_from_id
+
+
diff --git a/platform/micro-services/data-export/source/data_export/database/get_dataset_from_id.py b/platform/micro-services/data-export/source/data_export/database/get_dataset_from_id.py
new file mode 100644
index 00000000..2509e106
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/database/get_dataset_from_id.py
@@ -0,0 +1,10 @@
+
+def get_dataset_from_id(config, dataset_id):
+    print(config, config["datasets"].as_dict(), dataset_id)
+
+    if dataset_id in config["datasets"].as_dict():
+        return config["datasets"][dataset_id]
+
+    assert False, "Not implemented"
+
+
diff --git a/platform/micro-services/data-export/source/data_export/export/__init__.py b/platform/micro-services/data-export/source/data_export/export/__init__.py
new file mode 100644
index 00000000..327771a8
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/export/__init__.py
@@ -0,0 +1,4 @@
+
+from data_export.export.exporter_factory import ExporterFactory
+
+
diff --git a/platform/micro-services/data-export/source/data_export/export/cloud_exporter.py b/platform/micro-services/data-export/source/data_export/export/cloud_exporter.py
new file mode 100644
index 00000000..9fda1847
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/export/cloud_exporter.py
@@ -0,0 +1,31 @@
+
+import requests
+
+import logging
+
+logger = logging.getLogger()
+
+class CloudExporter:
+    def __init__(self, config):
+        self.config = config
+
+    def export(self, path, dataset):
+        url = self.config["exporter"]["endpoint"] + ":" + self.get_port() + "/peoples_speech/export_dataset"
+        data = {
+            "dataset" : dataset,
+            "output_dataset" : path
+        }
+
+        logger.debug("Submitting POST request to url " + str(url))
+        logger.debug(" with data " + str(data))
+        response = requests.post(url, json=data)
+
+        if response.status_code == 200:
+            print("Success")
+            print(response.json())
+        else:
+            print("Failed to submit with error: " + str(response.status_code))
+
+    def get_port(self):
+        return "5000"
+
diff --git a/platform/micro-services/data-export/source/data_export/export/exporter_factory.py b/platform/micro-services/data-export/source/data_export/export/exporter_factory.py
new file mode 100644
index 00000000..fcde7833
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/export/exporter_factory.py
@@ -0,0 +1,20 @@
+
+from data_export.export.google_cloud_parallel_exporter import GoogleCloudParallelExporter
+from data_export.export.local_exporter import LocalExporter
+from data_export.export.cloud_exporter import CloudExporter
+
+class ExporterFactory:
+    def __init__(self, config):
+        self.config = config
+
+    def create(self):
+        if self.config["exporter"]["type"] == "GoogleCloudParallelExporter":
+            return GoogleCloudParallelExporter(self.config)
+        if self.config["exporter"]["type"] == "LocalExporter":
+            return LocalExporter(self.config)
+        if self.config["exporter"]["type"] == "CloudExporter":
+            return CloudExporter(self.config)
+
+        assert False, "Unknown exporter type '" + self.config["exporter"]["type"] + "'"
+
+
diff --git a/platform/micro-services/data-export/source/data_export/export/google_cloud_parallel_exporter.py b/platform/micro-services/data-export/source/data_export/export/google_cloud_parallel_exporter.py
new file mode 100644
index 00000000..65a86574
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/export/google_cloud_parallel_exporter.py
@@ -0,0 +1,30 @@
+
+from data_export.google.google_cloud_work_queue import GoogleCloudWorkQueue
+import json
+
+class GoogleCloudParallelExporter:
+    def __init__(self, config):
+        self.config = config
+
+        self.work_queue = GoogleCloudWorkQueue(config)
+
+    def export(self, output_dataset_path, dataset):
+        print("Exporting dataset", dataset)
+
+        task_count = int(self.config["exporter"]["task_count"])
+
+        for task_id in range(task_count):
+            self.work_queue.push(json.dumps({
+                "task_id" : task_id,
+                "task_count" : task_count,
+                "dataset" : dataset}))
+
+        self.start_job()
+
+    def start_job(self):
+        kubernetes_directory = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(__file__)))), "build-scripts", "kubernetes")
+        command = "kubectl apply -f run-export-worker.yaml"
+        subprocess.run(command, cwd = kubernetes_directory)
+
+
+
diff --git a/platform/micro-services/data-export/source/data_export/export/local_exporter.py b/platform/micro-services/data-export/source/data_export/export/local_exporter.py
new file mode 100644
index 00000000..8a61b29e
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/export/local_exporter.py
@@ -0,0 +1,13 @@
+
+from data_export.utility.load_dataset_csv import load_dataset_csv
+from data_export.utility.write_samples_to_tar_gz import write_samples_to_tar_gz
+
+class LocalExporter:
+    def __init__(self, config):
+        self.config = config
+
+    def export(self, path, dataset):
+        samples = load_dataset_csv(dataset, task_id=0, task_count=1)
+
+        write_samples_to_tar_gz(samples, path)
+
diff --git a/platform/micro-services/data-export/source/data_export/google/export_worker.py b/platform/micro-services/data-export/source/data_export/google/export_worker.py
new file mode 100644
index 00000000..253fd51f
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/google/export_worker.py
@@ -0,0 +1,41 @@
+
+from data_export.google.google_cloud_work_queue import GoogleCloudWorkQueue
+from data_export.utility.get_config import get_config
+from data_export.utility.load_dataset_csv import load_dataset_csv
+from data_export.utility.write_samples_to_tar_gz import write_samples_to_tar_gz
+
+import json
+
+def main():
+    config = get_config()
+    q = GoogleCloudWorkQueue(config)
+    print("Worker with sessionID: " +  q.sessionID())
+    print("Initial queue state: empty=" + str(q.empty()))
+    while not q.empty():
+        item = q.lease(lease_secs=3600, block=True, timeout=2)
+        if item is not None:
+            itemstr = item.decode("utf-8")
+            print("Working on " + itemstr)
+            export_data(itemstr, config)
+            q.complete(item)
+        else:
+            print("Waiting for work")
+    print("Queue empty, exiting")
+
+def export_data(itemstr, config):
+    item = json.loads(itemstr)
+
+    task_id = item["task_id"]
+    task_count = item["task_count"]
+    dataset = item["dataset"]
+
+    samples = load_dataset_csv(dataset, task_id, task_count)
+
+    path = config["exporter"]["output_path"] + "-" + str(task_id) + "-tar.gz"
+
+    write_samples_to_tar_gz(samples, path)
+
+if __name__ == "__main__":
+    main()
+
+
diff --git a/platform/micro-services/data-export/source/data_export/google/google_cloud_work_queue.py b/platform/micro-services/data-export/source/data_export/google/google_cloud_work_queue.py
new file mode 100644
index 00000000..0b38a3b4
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/google/google_cloud_work_queue.py
@@ -0,0 +1,138 @@
+#!/usr/bin/env python
+
+# Based on http://peter-hoffmann.com/2012/python-simple-queue-redis-queue.html
+# and the suggestion in the redis documentation for RPOPLPUSH, at
+# http://redis.io/commands/rpoplpush, which suggests how to implement a work-queue.
+
+
+import redis
+import uuid
+import hashlib
+
+from data_export.google.redis_service import RedisService
+
+class GoogleCloudWorkQueue(object):
+    """Simple Finite Work Queue with Redis Backend
+
+    This work queue is finite: as long as no more work is added
+    after workers start, the workers can detect when the queue
+    is completely empty.
+
+    The items in the work queue are assumed to have unique values.
+
+    This object is not intended to be used by multiple threads
+    concurrently.
+    """
+    def __init__(self, config):
+        """The default connection parameters are: host='localhost', port=6379, db=0
+
+        The work queue is identified by "name".  The library may create other
+        keys with "name" as a prefix.
+        """
+
+        self.redis_service = RedisService(config)
+        self._db = redis.StrictRedis(**self.redis_service.kwargs())
+        # The session ID will uniquely identify this "worker".
+        self._session = str(uuid.uuid4())
+        # Work queue is implemented as two queues: main, and processing.
+        # Work is initially in main, and moved to processing when a client picks it up.
+        name = "exporter-work-queue"
+        self._main_q_key = name
+        self._processing_q_key = name + ":processing"
+        self._lease_key_prefix = name + ":leased_by_session:"
+
+    def push(self, item):
+        self._db.rpush(item)
+
+    def sessionID(self):
+        """Return the ID for this session."""
+        return self._session
+
+    def _main_qsize(self):
+        """Return the size of the main queue."""
+        return self._db.llen(self._main_q_key)
+
+    def _processing_qsize(self):
+        """Return the size of the main queue."""
+        return self._db.llen(self._processing_q_key)
+
+    def empty(self):
+        """Return True if the queue is empty, including work being done, False otherwise.
+
+        False does not necessarily mean that there is work available to work on right now,
+        """
+        return self._main_qsize() == 0 and self._processing_qsize() == 0
+
+# TODO: implement this
+#    def check_expired_leases(self):
+#        """Return to the work queueReturn True if the queue is empty, False otherwise."""
+#        # Processing list should not be _too_ long since it is approximately as long
+#        # as the number of active and recently active workers.
+#        processing = self._db.lrange(self._processing_q_key, 0, -1)
+#        for item in processing:
+#          # If the lease key is not present for an item (it expired or was
+#          # never created because the client crashed before creating it)
+#          # then move the item back to the main queue so others can work on it.
+#          if not self._lease_exists(item):
+#            TODO: transactionally move the key from processing queue to
+#            to main queue, while detecting if a new lease is created
+#            or if either queue is modified.
+
+    def _itemkey(self, item):
+        """Returns a string that uniquely identifies an item (bytes)."""
+        return hashlib.sha224(item).hexdigest()
+
+    def _lease_exists(self, item):
+        """True if a lease on 'item' exists."""
+        return self._db.exists(self._lease_key_prefix + self._itemkey(item))
+
+    def lease(self, lease_secs=60, block=True, timeout=None):
+        """Begin working on an item the work queue.
+
+        Lease the item for lease_secs.  After that time, other
+        workers may consider this client to have crashed or stalled
+        and pick up the item instead.
+
+        If optional args block is true and timeout is None (the default), block
+        if necessary until an item is available."""
+        if block:
+            item = self._db.brpoplpush(self._main_q_key, self._processing_q_key, timeout=timeout)
+        else:
+            item = self._db.rpoplpush(self._main_q_key, self._processing_q_key)
+        if item:
+            # Record that we (this session id) are working on a key.  Expire that
+            # note after the lease timeout.
+            # Note: if we crash at this line of the program, then GC will see no lease
+            # for this item a later return it to the main queue.
+            itemkey = self._itemkey(item)
+            self._db.setex(self._lease_key_prefix + itemkey, lease_secs, self._session)
+        return item
+
+    def complete(self, value):
+        """Complete working on the item with 'value'.
+
+        If the lease expired, the item may not have completed, and some
+        other worker may have picked it up.  There is no indication
+        of what happened.
+        """
+        self._db.lrem(self._processing_q_key, 0, value)
+        # If we crash here, then the GC code will try to move the value, but it will
+        # not be here, which is fine.  So this does not need to be a transaction.
+        itemkey = self._itemkey(value)
+        self._db.delete(self._lease_key_prefix + itemkey)
+
+# TODO: add functions to clean up all keys associated with "name" when
+# processing is complete.
+
+# TODO: add a function to add an item to the queue.  Atomically
+# check if the queue is empty and if so fail to add the item
+# since other workers might think work is done and be in the process
+# of exiting.
+
+# TODO(etune): move to my own github for hosting, e.g. github.com/erictune/rediswq-py and
+# make it so it can be pip installed by anyone (see
+# http://stackoverflow.com/questions/8247605/configuring-so-that-pip-install-can-work-from-github)
+
+# TODO(etune): finish code to GC expired leases, and call periodically
+#  e.g. each time lease times out.
+
diff --git a/platform/micro-services/data-export/source/data_export/google/redis_service.py b/platform/micro-services/data-export/source/data_export/google/redis_service.py
new file mode 100644
index 00000000..d48abb98
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/google/redis_service.py
@@ -0,0 +1,51 @@
+
+import subprocess
+import logging
+
+logger = logging.getLogger(__name__)
+
+class RedisService:
+    def __init__(self, config):
+        self.config = config
+
+        self.startup()
+
+    def startup(self):
+        self.instances = self.get_instances()
+
+        if not "exporter-work-queue" in self.instances:
+            self.start_instance()
+
+    def start_instance(self):
+        command = ["gcloud", "redis", "instances", "create", "exporter-work-queue", "--size=2", "--region", "europe-west4"]
+
+        subprocess.run(command)
+
+    def get_instances(self):
+        command = ["gcloud", "redis", "instances", "describe", "exporter-work-queue", "--region", "europe-west4"]
+
+        process = subprocess.run(command, capture_output=True)
+
+        stdout = process.stdout.decode()
+
+        logger.debug("stdout: " + stdout)
+
+        # parse output
+        instances = {}
+
+        for line in stdout.split('\n'):
+            logger.debug("parsing line: '" + line + "'")
+            if line.find("host:") == 0:
+                host = line[5:].strip()
+
+            if line.find("name:") == 0:
+                name = line[5:].strip().split("/")[-1]
+                instances[name] = host
+
+        return instances
+
+    def kwargs(self):
+        return { "host" : self.instances["exporter-work-queue"], "port" : 6379, "db" : 0 }
+
+
+
diff --git a/platform/micro-services/data-export/source/data_export/utility/__init__.py b/platform/micro-services/data-export/source/data_export/utility/__init__.py
new file mode 100644
index 00000000..1da8520e
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/utility/__init__.py
@@ -0,0 +1,4 @@
+
+from data_export.utility.get_config import get_config
+
+
diff --git a/platform/micro-services/data-export/source/data_export/utility/get_config.py b/platform/micro-services/data-export/source/data_export/utility/get_config.py
new file mode 100644
index 00000000..375d4af2
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/utility/get_config.py
@@ -0,0 +1,47 @@
+
+import config
+import os
+import logging
+
+def get_config():
+    config = setup_config({})
+
+    setup_logging(config)
+
+    return config
+
+def setup_config(dictionary):
+    return config.ConfigurationSet(
+        config.config_from_env(prefix="MLCOMMONS"),
+        config.config_from_yaml(config_path(), read_from_file=True),
+        config.config_from_dict(dictionary),
+    )
+
+def config_path():
+    home = os.path.expanduser("~")
+    home_config_path = os.path.join(home, ".mlcommons", "config.yaml")
+    if os.path.exists(home_config_path):
+        return home_config_path
+
+    return os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(__file__))), "configs", "server.yaml")
+
+def setup_logging(arguments):
+
+    logging_format = "%(asctime)s - %(levelname)s - %(name)s - %(message)s"
+
+    if arguments["verbose"]:
+        logging.basicConfig(level=logging.DEBUG, format=logging_format)
+    elif arguments["verbose_info"]:
+        logging.basicConfig(level=logging.INFO, format=logging_format)
+    else:
+        logging.basicConfig(level=logging.WARNING, format=logging_format)
+
+    root_logger = logging.getLogger()
+
+    if arguments["verbose"]:
+        root_logger.setLevel(logging.DEBUG)
+    elif arguments["verbose_info"]:
+        root_logger.setLevel(logging.INFO)
+    else:
+        root_logger.setLevel(logging.WARNING)
+
diff --git a/platform/micro-services/data-export/source/data_export/utility/load_dataset_csv.py b/platform/micro-services/data-export/source/data_export/utility/load_dataset_csv.py
new file mode 100644
index 00000000..3d4e0856
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/utility/load_dataset_csv.py
@@ -0,0 +1,31 @@
+
+from smart_open import open
+
+import csv
+
+import json
+
+import logging
+
+logging.getLogger("urllib3").setLevel(logging.WARNING)
+logging.getLogger("google").setLevel(logging.WARNING)
+
+def load_dataset_csv(csv_path, task_id, task_count):
+
+    with open(csv_path) as csv_file:
+        return (yield from load_dataset_csv_from_file(csv_file, task_id, task_count))
+
+def load_dataset_csv_from_file(csv_file, task_id, task_count):
+        reader = csv.reader(csv_file, delimiter=',', quotechar='"')
+        index = 0
+        for row in reader:
+            path, caption = row[0], row[1]
+
+            metadata = {}
+            if len(row) >= 3:
+                metadata = json.loads(row[2])
+
+            if index % task_count == task_id:
+                yield {"path" : path, "caption" : caption, "metadata" : metadata}
+
+            index += 1
diff --git a/platform/micro-services/data-export/source/data_export/utility/write_samples_to_tar_gz.py b/platform/micro-services/data-export/source/data_export/utility/write_samples_to_tar_gz.py
new file mode 100644
index 00000000..b407ade6
--- /dev/null
+++ b/platform/micro-services/data-export/source/data_export/utility/write_samples_to_tar_gz.py
@@ -0,0 +1,51 @@
+
+import os
+import shutil
+import subprocess
+import csv
+import json
+import errno
+
+def write_samples_to_tar_gz(samples, path):
+    os.makedirs("/tmp/export-data", exist_ok=True)
+
+    # copy locally
+    command = ["gsutil", "-m", "cp", "-I", "/tmp/export-data"]
+
+    stream_to_command(command, samples)
+
+    # tar it
+    filename = os.path.basename(path)
+
+    temp_archive = os.path.join("/tmp", filename)
+
+    command = ["tar", "-czvf", temp_archive, "export-data"]
+
+    subprocess.run(command, cwd="/tmp")
+
+    # upload it
+    command = ["gsutil", "-m", "cp", temp_archive, path]
+
+    subprocess.run(command)
+
+def stream_to_command(command, samples):
+    p = subprocess.Popen(command, stdin=subprocess.PIPE)
+    with open("/tmp/export-data/dataset.csv", "w", newline="") as csv_file:
+        writer = csv.writer(csv_file, delimiter=',', quotechar='"')
+
+        for sample in samples:
+            writer.writerow([sample["path"], sample["caption"], json.dumps(sample["metadata"])])
+            line = sample["path"] + "\n"
+            try:
+                p.stdin.write(line.encode('utf-8'))
+            except IOError as e:
+                if e.errno == errno.EPIPE or e.errno == errno.EINVAL:
+                    # Stop loop on "Invalid pipe" or "Invalid argument".
+                    # No sense in continuing with broken pipe.
+                    break
+                else:
+                    # Raise any other error.
+                    raise
+
+    p.stdin.close()
+    p.wait()
diff --git a/platform/micro-services/data-export/source/flask/app.py b/platform/micro-services/data-export/source/flask/app.py
new file mode 100644
index 00000000..289198fb
--- /dev/null
+++ b/platform/micro-services/data-export/source/flask/app.py
@@ -0,0 +1,18 @@
+from flask import Flask, request
+from flask_cors import CORS #comment this on deployment
+
+import data_export
+
+import logging
+
+app = Flask(__name__)
+CORS(app) #comment this on deployment
+
+logger = logging.getLogger(__name__)
+
+@app.route('/peoples_speech/export_dataset', methods=['GET', 'POST'])
+def export_dataset():
+    dataset = request.json["dataset"]
+    output_dataset = request.json["output_dataset"]
+    data_export.export_dataset(output_dataset, dataset)
+    return { "results_path" : output_dataset }
diff --git a/platform/micro-services/data-export/start-dev b/platform/micro-services/data-export/start-dev
new file mode 100755
index 00000000..31eabc6b
--- /dev/null
+++ b/platform/micro-services/data-export/start-dev
@@ -0,0 +1,51 @@
+#! /bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# u unset variables are errors
+# f disable globbing on *
+# pipefail | produces a failure code if any stage fails
+set -euf -o pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+# Setup virtual environment
+PYTHON_ENV=$(python3 -c "import sys; sys.stdout.write(sys.prefix) if (hasattr(sys, 'real_prefix') or sys.base_prefix != sys.prefix) else sys.stdout.write('0')")
+if [[ $PYTHON_ENV == 0 ]];
+then
+echo "Not in virtual environment"
+
+ACTIVATE=$LOCAL_DIRECTORY/environment/bin/activate
+
+if [ ! -f $ACTIVATE ]; then
+echo "Virtual environment doesn't exist, making it..."
+python3 -m venv $LOCAL_DIRECTORY/environment
+python3 -m pip install --upgrade pip > /dev/null
+fi
+
+source $ACTIVATE
+else
+echo "Running in virtual environment $PYTHON_ENV"
+fi
+
+# Make sure requirements are installed
+pip install -r $LOCAL_DIRECTORY/requirements.txt > /dev/null
+
+# Set python environment
+PYTHONPATH="$LOCAL_DIRECTORY/source/flask"
+PYTHONPATH+=":$LOCAL_DIRECTORY/source"
+export PYTHONPATH
+
+# export flask
+export FLASK_APP=$LOCAL_DIRECTORY/source/flask/app.py
+export FLASK_DEBUG=1
+
+# kill background tasks on script exit
+trap "trap - SIGTERM && kill -- -$$" SIGINT SIGTERM EXIT
+
+# Start the dev environment
+python -m flask run &
+wait
+
+
diff --git a/platform/micro-services/data-export/start-production b/platform/micro-services/data-export/start-production
new file mode 100755
index 00000000..1c653dd0
--- /dev/null
+++ b/platform/micro-services/data-export/start-production
@@ -0,0 +1,51 @@
+#! /bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# u unset variables are errors
+# f disable globbing on *
+# pipefail | produces a failure code if any stage fails
+set -euf -o pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+# Setup virtual environment
+PYTHON_ENV=$(python3 -c "import sys; sys.stdout.write(sys.prefix) if (hasattr(sys, 'real_prefix') or sys.base_prefix != sys.prefix) else sys.stdout.write('0')")
+if [[ $PYTHON_ENV == 0 ]];
+then
+echo "Not in virtual environment"
+
+ACTIVATE=$LOCAL_DIRECTORY/environment/bin/activate
+
+if [ ! -f $ACTIVATE ]; then
+echo "Virtual environment doesn't exist, making it..."
+python3 -m venv $LOCAL_DIRECTORY/environment
+python3 -m pip install --upgrade pip > /dev/null
+fi
+
+source $ACTIVATE
+else
+echo "Running in virtual environment $PYTHON_ENV"
+fi
+
+# Make sure requirements are installed
+pip install -r $LOCAL_DIRECTORY/requirements.txt > /dev/null
+
+# Set python environment
+PYTHONPATH="$LOCAL_DIRECTORY/source/flask"
+PYTHONPATH+=":$LOCAL_DIRECTORY/source"
+export PYTHONPATH
+
+# export flask
+export FLASK_APP=$LOCAL_DIRECTORY/source/flask/app.py
+export FLASK_DEBUG=0
+
+# kill background tasks on script exit
+trap "trap - SIGTERM && kill -- -$$" SIGINT SIGTERM EXIT
+
+# Start the dev environment
+python -m flask run --host=0.0.0.0 --port=$1 &
+wait
+
+
diff --git a/platform/micro-services/peoples-speech/source/peoples_speech/data_export/__init__.py b/platform/micro-services/peoples-speech/source/peoples_speech/data_export/__init__.py
index ffea867e..ffdb4f4b 100644
--- a/platform/micro-services/peoples-speech/source/peoples_speech/data_export/__init__.py
+++ b/platform/micro-services/peoples-speech/source/peoples_speech/data_export/__init__.py
@@ -1,3 +1,4 @@
 
 from data_export.api.save_dataset import save_dataset
+from data_export.api.setup_cli_parser import setup_cli_parser
 
diff --git a/platform/micro-services/quality/scripts/sample-dataset.py b/platform/micro-services/quality/scripts/sample-dataset.py
new file mode 100644
index 00000000..4444b722
--- /dev/null
+++ b/platform/micro-services/quality/scripts/sample-dataset.py
@@ -0,0 +1,131 @@
+
+import os
+
+from argparse import ArgumentParser
+
+import config
+import jsonlines
+import random
+from smart_open import open
+
+import logging
+
+logger = logging.getLogger(__name__)
+
+def main():
+    parser = ArgumentParser("Randomly sample audio and transcripts from a json lines file.")
+
+    subparsers = parser.add_subparsers()
+
+    setup_cli_parser(subparsers)
+
+    args = parser.parse_args()
+
+    args.func(args)
+
+def setup_cli_parser(subparsers):
+
+    parser = subparsers.add_parser('dataset')
+
+    parser.add_argument("-i", "--dataset-path",
+        default="gs://the-peoples-speech-west-europe/forced-aligner/cuda-forced-aligner/output_work_dir_5b/output_work_dir_5b/dataset_manifest_mp3_956_all.json",
+        help="Path to json lines file describing the dataset.")
+    parser.add_argument("-o", "--output-dataset-path", default="data/dataset.json", help="The path to save the new dataset.")
+    parser.add_argument("-c", "--config-file-path", default="", help="The path to the config file.")
+    parser.add_argument("-m", "--maximum-samples", default=100, help="How many samples to download.")
+    parser.add_argument("--maximum-dataset-size", default=1000, help="How many samples to scan.")
+    parser.add_argument("-v", "--verbose", default=False, action="store_true", help="Print out debug messages.")
+    parser.add_argument("-vi", "--verbose-info", default=False, action="store_true", help="Print out info messages.")
+
+    parser.set_defaults(func=dispatch)
+
+def dispatch(args):
+    arguments = vars(args)
+
+    config = setup_config(arguments)
+
+    setup_logging(config)
+
+    logger.debug("Full config: " + str(config))
+
+    samples = load_samples(config)
+
+    filtered_samples = filter_samples(samples, config)
+
+    save_samples(filtered_samples, config)
+
+def load_samples(config):
+    samples = []
+
+    with open(config["dataset_path"]) as dataset_file:
+
+        dataset_reader = jsonlines.Reader(dataset_file)
+
+        for line in dataset_reader:
+            for path, label in zip(line["training_data"]["labels"], line["training_data"]["output_paths"]):
+                samples.append(path, label)
+
+            if len(samples) > int(config["maximum_dataset_size"]):
+                break
+
+    return samples
+
+def filter_samples(samples, config):
+    random.seed(42)
+    random.shuffle(samples)
+
+    limit = min(len(samples), config["maximum_samples"])
+
+    return samples[:limit]
+
+def save_samples(samples, config):
+
+    output_directory = os.path.dirname(config["output_dataset_path"])
+    os.mkdirs(output_directory, exist_ok=True)
+
+    with open(config["output_dataset_path"]) as dataset_file:
+        dataset_writer = jsonlines.Writer(dataset_file)
+
+        for path, label in samples:
+            local_path = download(path, output_directory)
+            dataset_writer({"training_datset" : { "output_paths": [local_path], "labels" : [label] }})
+
+def setup_config(dictionary):
+    return config.ConfigurationSet(
+        config.config_from_env(prefix="MLCOMMONS"),
+        config.config_from_yaml(config_path(), read_from_file=True),
+        config.config_from_dict(dictionary),
+    )
+
+def config_path():
+    home = os.path.expanduser("~")
+    home_config_path = os.path.join(home, ".mlcommons", "config.yaml")
+    if os.path.exists(home_config_path):
+        return home_config_path
+
+    return os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(__file__))), "config", "default.yaml")
+
+def setup_logging(arguments):
+
+    logging_format = "%(asctime)s - %(levelname)s - %(name)s - %(message)s"
+
+    if arguments["verbose"]:
+        logging.basicConfig(level=logging.DEBUG, format=logging_format)
+    elif arguments["verbose_info"]:
+        logging.basicConfig(level=logging.INFO, format=logging_format)
+    else:
+        logging.basicConfig(level=logging.WARNING, format=logging_format)
+
+    root_logger = logging.getLogger()
+
+    if arguments["verbose"]:
+        root_logger.setLevel(logging.DEBUG)
+    elif arguments["verbose_info"]:
+        root_logger.setLevel(logging.INFO)
+    else:
+        root_logger.setLevel(logging.WARNING)
+
+if __name__ == "__main__":
+    main()
+
+
diff --git a/platform/micro-services/website/build-scripts/cloud-build/cloudbuild.sh b/platform/micro-services/website/build-scripts/cloud-build/cloudbuild.sh
new file mode 100755
index 00000000..adab36dc
--- /dev/null
+++ b/platform/micro-services/website/build-scripts/cloud-build/cloudbuild.sh
@@ -0,0 +1 @@
+gcloud builds submit --config cloudbuild.yaml ../..
diff --git a/platform/micro-services/website/build-scripts/cloud-build/cloudbuild.yaml b/platform/micro-services/website/build-scripts/cloud-build/cloudbuild.yaml
new file mode 100644
index 00000000..00ca2d40
--- /dev/null
+++ b/platform/micro-services/website/build-scripts/cloud-build/cloudbuild.yaml
@@ -0,0 +1,3 @@
+steps:
+- name: 'gcr.io/cloud-builders/docker'
+  args: [ 'build', '-t', 'gcr.io/peoples-speech/data-export', '-f', 'build-scripts/docker/Dockerfile', '.' ]
diff --git a/platform/micro-services/website/build-scripts/docker/Dockerfile b/platform/micro-services/website/build-scripts/docker/Dockerfile
new file mode 100644
index 00000000..76fcae76
--- /dev/null
+++ b/platform/micro-services/website/build-scripts/docker/Dockerfile
@@ -0,0 +1,8 @@
+FROM ubuntu:20.04
+
+COPY . /app
+
+RUN apt-get update && apt-get install -y python3-pip && pip3 install -r /app/requirements.txt
+
+CMD /app/start-production
+
diff --git a/platform/micro-services/website/build-scripts/docker/build-container.sh b/platform/micro-services/website/build-scripts/docker/build-container.sh
new file mode 100755
index 00000000..561a0400
--- /dev/null
+++ b/platform/micro-services/website/build-scripts/docker/build-container.sh
@@ -0,0 +1,14 @@
+#! /bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# u unset variables are errors
+# f disable globbing on *
+# pipefail | produces a failure code if any stage fails
+set -euf -o pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+docker build -t peoples-speech-platform:0.1 -f $LOCAL_DIRECTORY/Dockerfile $LOCAL_DIRECTORY/../..
+
diff --git a/platform/micro-services/website/build-scripts/kubernetes/deploy-staging.sh b/platform/micro-services/website/build-scripts/kubernetes/deploy-staging.sh
new file mode 100644
index 00000000..0254b043
--- /dev/null
+++ b/platform/micro-services/website/build-scripts/kubernetes/deploy-staging.sh
@@ -0,0 +1,2 @@
+gcloud container clusters get-credentials peoples-speech-platform
+kubectl create deployment website --image=gcr.io/peoples-speech/platform:latest
diff --git a/platform/micro-services/website/start-production b/platform/micro-services/website/start-production
new file mode 100755
index 00000000..cea26de8
--- /dev/null
+++ b/platform/micro-services/website/start-production
@@ -0,0 +1,66 @@
+#! /bin/bash
+
+# Safely execute this bash script
+# e exit on first failure
+# u unset variables are errors
+# f disable globbing on *
+# pipefail | produces a failure code if any stage fails
+set -euf -o pipefail
+
+# Get the directory of this script
+LOCAL_DIRECTORY="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+# Setup virtual environment
+PYTHON_ENV=$(python3 -c "import sys; sys.stdout.write(sys.prefix) if (hasattr(sys, 'real_prefix') or sys.base_prefix != sys.prefix) else sys.stdout.write('0')")
+if [[ $PYTHON_ENV == 0 ]];
+then
+echo "Not in virtual environment"
+
+ACTIVATE=$LOCAL_DIRECTORY/environment/bin/activate
+
+if [ ! -f $ACTIVATE ]; then
+echo "Virtual environment doesn't exist, making it..."
+python3 -m venv $LOCAL_DIRECTORY/environment
+python3 -m pip install --upgrade pip > /dev/null
+fi
+
+source $ACTIVATE
+else
+echo "Running in virtual environment $PYTHON_ENV"
+fi
+
+# Make sure requirements are installed
+pip install -r $LOCAL_DIRECTORY/requirements.txt > /dev/null
+
+# Set python environment
+PYTHONPATH="$LOCAL_DIRECTORY/source/flask"
+export PYTHONPATH
+
+# Setup google credentials
+export GOOGLE_APPLICATION_CREDENTIALS=$LOCAL_DIRECTORY/source/config/gcloud_key.json
+
+# export flask
+export FLASK_APP=$LOCAL_DIRECTORY/source/flask/app.py
+
+# Setup the react dev environment
+REACT_DEV_ENVIRONMENT=$LOCAL_DIRECTORY/react-development
+
+# Setup the dev environment if it doesn't exist
+if [ ! -d $REACT_DEV_ENVIRONMENT ]; then
+npx create-react-app $REACT_DEV_ENVIRONMENT
+fi
+
+# Move the code over
+rsync -av --delete $LOCAL_DIRECTORY/source/react/ $REACT_DEV_ENVIRONMENT/src/
+rsync -av --delete $LOCAL_DIRECTORY/source/config/ $REACT_DEV_ENVIRONMENT/src/config/
+
+# kill background tasks on script exit
+trap "trap - SIGTERM && kill -- -$$" SIGINT SIGTERM EXIT
+
+# Start the dev environment
+cd $REACT_DEV_ENVIRONMENT
+python -m flask run &
+COLOR=1 npm start | cat &
+wait
+
+
diff --git a/platform/peoples-speech b/platform/peoples-speech
index 437320a7..af1105e3 100755
--- a/platform/peoples-speech
+++ b/platform/peoples-speech
@@ -16,11 +16,11 @@ if [[ $PYTHON_ENV == 0 ]];
 then
 echo "Not in virtual environment"
 
-ACTIVATE=$LOCAL_DIRECTORY/environment/bin/activate
+ACTIVATE=/tmp/mlcommons-environment/bin/activate
 
 if [ ! -f $ACTIVATE ]; then
 echo "Virtual environment doesn't exist, making it..."
-python3 -m venv $LOCAL_DIRECTORY/environment
+python3 -m venv /tmp/mlcommons-environment
 python -m pip install --upgrade pip > /dev/null
 fi
 
diff --git a/platform/requirements.txt b/platform/requirements.txt
index fd3a245b..42020b04 100644
--- a/platform/requirements.txt
+++ b/platform/requirements.txt
@@ -4,3 +4,4 @@ gtts
 python-configuration[yaml]
 whoosh
 audiofile
+redis