Demand for spoken audio content such as podcasts remains robust despite the proliferation of video services and other entertainment options for consumers. Shibin Li, Co-founder of Castbox, credits growth of the global podcast platform to the following: speed and availability, market-leading features, the proliferation of smart devices to deliver audio content, and activities — such as driving and working around the house — that make consuming video difficult.

Founded in 2016 and headquartered in Beijing, China, Castbox enables users to locate, access, and create spoken audio content. Available on iOS and Android, Castbox supports 50 million podcasts, on-demand radio programs, and audiobooks in 70 languages from 175 countries. The platform hosts about 2 million users per day and is the largest podcast platform on Android.

Castbox includes a range of features that build on its core service to provide a high-quality user experience. These features include curated podcast recommendations and in-audio search.

A combination of services

At its inception, Castbox relied on a combination of a multinational cloud services, as well as Google Cloud Platform (GCP) services, including the Google BigQuery analytics data warehouse and Cloud APIs to provide programmatic interfaces with Google Cloud services, and a range of services from the Google mobile development platform Firebase.

However, as Castbox matured and its user base expanded, the business increased its reliance on Google Cloud Platform and Firebase.

“We needed to access stable cloud services as we could not tolerate long periods of downtime that would compromise the user experience,” says Li. “Furthermore, we had to support up to 50,000 concurrent connections, and potentially more in future, without disruption.”

“Based on our analysis of the data in Google BigQuery, we can determine what type of content users are listening to, how long they like to listen to it, and when they like to listen to it. This allows us to recommend similar podcasts to each user based on the preferences he or she expressed, encouraging activity on and return visits to our platform.”

—Shibin Li, Co-Founder, Castbox

Competitive differentiation

Castbox also found machine learning-powered Google Cloud Platform APIs could help deliver features, such as in-audio search, that differentiate the podcast platform from its competitors. In addition, Firebase SDKs and Firebase A/B Testing would enable Castbox to create and analyze new applications, as well as make adjustments based on user feedback. Firebase Realtime Database would allow the business to support tens of thousands of concurrent user connections.

The diligence of the Google Cloud team in advising Li and her team about forthcoming products and services also swayed Castbox towards Google technologies. The business gained the opportunity with Google to join several programs that offered early access to Google innovations.

Signature in-audio search service

Castbox now uses Google Cloud Platform services in the Tokyo, Japan, and U.S. East regions. Cloud Speech-to-Text API plays a key role in delivering Castbox’s signature in-audio search service. This service enables users to search transcriptions of audio content on the platform for words or phrases. The search results incorporate the title of the podcast and the search term in context (for example, within the sentence or sentence excerpt in which it appears). Each use of the word or phrase is time-stamped so it can easily be found. The API enables Castbox developers to apply neural network algorithms to achieve audio-to-text conversion accuracy rates of greater than 96%, while search queries typically experience latency of just 50 milliseconds.

In addition, the latency of comparison data, converting audio to text, is only about 250 milliseconds, contributing to the processing of about 12 minutes worth of audio to text in just 10 minutes. “We can process about 20 hours of audio files in one day,” Li says. “This enables us to transcribe and index all the new episodes of a podcast in that period.”

50,000 concurrent connections

With Firebase Realtime Database, Castbox now supports up to 50,000 concurrent connections to its platform with an average latency per connection of just 10 milliseconds. “Firebase Realtime Database also allows us to continue operating in offline mode, which is extremely helpful if we experience any network disruptions,” Li explains. “When we come back online again, any data is simply synchronized with the database.”

Google BigQuery and the analytics capabilities of Firebase SDKs also enable Castbox to monitor and analyze user behaviors. “Based on our analysis of the data in Google BigQuery, we can determine what type of content users are listening to, how long they like to listen to it, and when they like to listen to it,” Li explains. “This allows us to recommend similar podcasts to each user based on the preferences he or she expressed, encouraging activity on and return visits to our platform.”

“Given our queries may span up to 40 days of data, we may be analyzing up to 1,200 GB at one time. We have no problem doing this with Google BigQuery.”

—Shibin Li, Co-Founder, Castbox

Castbox also uses its analyses of Google BigQuery data to amend banners and summaries on its platform to encourage users to listen to additional content. Furthermore, the service is prepared to make surprise recommendations of content to users based on the preferences and reactions of users with similar tastes.

“We analyze a pool of data growing at up to 30 GB per day,” Li explains. “Given our queries may span up to 40 days of data, we may be analyzing up to 1,200 GB at one time. We have no problem doing this with Google BigQuery.” These analyses also support Castbox’s decision to start creating original content, such as finance and economic news, for its platform.

Checking weekly changes

Castbox does not rely only on analyzing user data to deliver a high-quality experience. The business aggregates user feedback from emails and Google Play reviews to make weekly changes to its platform. It then uses Firebase A/B Testing to check whether these changes are met with a positive user response.

“We are extremely pleased with Google Cloud Platform and Firebase. We have been able to differentiate ourselves from our competitors and provide an attractive option for users at a time when content and entertainment options are exploding. We have a great opportunity with Google to continue to improve the value of our offering to users and build engagement and loyalty.”

—Shibin Li, Co-Founder, Castbox

Castbox’s positive experiences with Google Cloud Platform are encouraging the business to grow its use of the product. “We are trying to move some more services to Google Cloud Platform because it is very stable and scalable,” Li says. The business is keen to explore the capabilities of Cloud Pub/Sub to provide low latency messaging between applications, Cloud Spanner to deliver a distributed relational database service, and Cloud Dataflow to transform and enrich data in stream and batch modes.

“We are extremely pleased with Google Cloud Platform and Firebase,” Li concludes. “We have been able to differentiate ourselves from our competitors and provide an attractive option for users at a time when content and entertainment options are exploding. We have a great opportunity with Google to continue to improve the value of our offering to users and build engagement and loyalty.”

This post shows how to build an image search utility using natural language queries. Our aim is to use different GCP services to demonstrate this. At the core of our project is OpenAI’s CLIP model. It makes use of two encoders – one for images and one for texts. Each encoder is trained to learn representations such that similar images and text embeddings are projected as close as possible.

We will first create a Flask-based REST API capable of handling natural language queries and matching them against relevant images. We will then demonstrate the use of the API through a Flutter-based web and mobile application. Figure 1 shows how our final application would look like:

All the code shown in this post is available as a GitHub repository. Let’s dive in. 

Application at a high-level

Our application will take two queries from the user:

• Tag or keyword query. This is needed in order to pull a set of images of interest from Pixabay. You can use any other image repositories for this purpose. But we found Pixabay’s API to be easier to work with. We will cache these images to optimize the user experience. Suppose we wanted to find images that are similar to this query: “horses amidst flowers”. For this, we’d first pull in a few “horse” images and then run another utility to find out the images that best match our query.  
• Longer or semantic query that we will use to retrieve the images from the pool created in the step above. These images should be semantically similar to this query. 

Note: Instead of two queries, we could have only taken a single long query and run named-entity extraction to determine the most likely important keywords to run the initial search with. For this post, we won’t be using this approach. 

Figure 2 below depicts the architecture design of our application and the technical stack used for each of the components.

Figure 2 also presents the core logic of the API we will develop in bits and pieces in this post. We will deploy this API on a Kubernetes cluster using the Google Kubernetes Engine (GKE). The following presents a brief directory structure of our application code-base:

Next, we will walk through the code and other related components for building our image search API. For various machine learning-related utilities, we will be using PyTorch. 

Building the backend API with Flask

First, we’d need to fetch a set of images with respect to user-provided tags/keywords before performing the natural language image search. The utility below from the pixabay_utils.py script can do this for us:

  def fetch_images_tag(pixabay_search_keyword, num_images):
    """
    Fetches images from Pixabay w.r.t a keyword.
    :param pixabay_search_keyword: Keyword to perform the search on Pixabay.
    :param num_images: Number of images to retrieve.
    :return: List of PIL images.
    :return: List of image URLs.
    """
    query = (
        PIXABAY_API
        + "&q="
        + pixabay_search_keyword.lower()
        + "&image_type=photo&safesearch=true&per_page="
        + str(num_images)
    )
    
    response = requests.get(query)
    output = response.json()

    all_images = []
    all_image_urls = []

    for each in output["hits"]:
        imageurl = each["webformatURL"]
        response = requests.get(imageurl)
        image = Image.open(BytesIO(response.content)).convert("RGB")
        all_images.append(image)
        all_image_urls.append(imageurl)

    return (all_images, all_image_urls)

Note that all the API utilities are logging relevant information. But for brevity, we have omitted the lines of code responsible for that. Next, we will see how to invoke the CLIP model and select the images that would best match a given query semantically.  For this, we’ll be using Hugging Face, an easy-to-use Python library offering state-of-the-art NLP capabilities. We’ll collate all the logic related to this search inside a SimilarityUtil class:

  class SimilarityUtil:
    def __init__(self):
        self.model = CLIPModel.from_pretrained(CLIP_MODEL)
        self.processor = CLIPProcessor.from_pretrained(CLIP_PREPROCESSOR)
        self.device = "cuda" if torch.cuda.is_available() else "cpu"

    def perform_sim_search(self, images, query_phrase, top_k=3):
        """
        Performs similarity search between the images and query.
        :param images: A list of PIL images initially retrieved with
        respect to some entity e.g. Tiger.
        :param query_phrase: A list containing a single text query,
        e.g. "Tiger drinking water".
        :param top_k: Number of top images to return from `images`.
        :return: Top-k indices matching the query semantically and
        their similarity scores.
        """
        model = self.model.to(self.device)
        # Obtain the text-image similarity scores
        with torch.no_grad():
            inputs = self.processor(
                text=[query_phrase], images=images, return_tensors="pt", padding=True
            )
            inputs = inputs.to(self.device)
            outputs = model(**inputs)

        # Image-text similarity scores
        logits_per_image = outputs.logits_per_image.cpu()
        (top_indices, top_scores) = self.sort_scores(logits_per_image, top_k)

        return (top_indices, top_scores)

    def sort_scores(self, scores, top_k):
        """
        Sorts the scores in a descending manner.
        :param scores: Scores to sort through.
        :param top_k: Number of top scores to return.
        :return: Top-k scores and their indices.
        """
        values, indices = scores.squeeze().topk(top_k)
        top_indices, top_scores = [], []

        for score, index in zip(values, indices):
            top_indices.append(int(index.numpy()))
            score = score.numpy().tolist()
            top_scores.append(round(score, 3))

        return (top_indices, top_scores)

CLIP_MODEL uses a ViT-base model to encode the images for generating meaningful embeddings with respect to the provided query. The text-based query is also encoded using A Transformers-based model for generating the embeddings. These two embeddings are matched with one another during inference. To know more about the particular methods we are using for the CLIP model please refer to this documentation from Hugging Face. 

In the code above, we are first invoking the CLIP model with images and the natural language query. This gives us a vector (logits_per_image) that contains the similarity scores between each of the images and the query. We then sort the vector in a descending manner. Note that we are initializing the CLIP model while instantiating the SimilarityUtil to save us the model loading time. This is the meat of our application and we have tackled it already. If you want to interact with this utility in a live manner you can check out this Colab Notebook. 

Now, we need to collate our utilities for fetching images from Pixabay and for performing the natural language image search inside a single script – perform_search.py. Following is the main class of that script:

  class Searcher:
    def __init__(self):
        self.similarity_model = SimilarityUtil()

    def get_similar_images(self, keyword, semantic_query, pixabay_max, top_k):
        """
        Finds semantically similar images.
        :param keyword: Keyword to search with on Pixabay.
        :param semantic_query: Query to find semantically similar images retrieved from Pixabay.
        :param pixabay_max: Number of maximum images to retrieve from Pixabay.
        :param top_k: Top-k images to return.
        :return: Tuple of top_k URLs and the similarity scores of the images present inside the URLs.
        """
        images_redis_key = keyword + "_images"
        urls_redis_key = keyword + "_urls"

        if redis_client.exists(images_redis_key) and redis_client.exists(
            urls_redis_key
        ):
            keyword_images = redis_client.get(images_redis_key)
            keyword_image_urls = redis_client.get(urls_redis_key)
        else:
            (keyword_images, keyword_image_urls) = fetch_images_tag(
                keyword, pixabay_max
            )
            redis_client.set(images_redis_key, keyword_images)
            redis_client.set(urls_redis_key, keyword_image_urls)

        (top_indices, top_scores) = self.similarity_model.perform_sim_search(
            keyword_images, semantic_query, top_k
        )

        top_urls = [keyword_image_urls[index] for index in top_indices]

        return (top_urls, top_scores)

Here, we are just calling the utilities we had previously developed to return the URLs of the most similar images and their scores. What is even more important here is the caching capability. For that, we combined GCP’s MemoryStore and a Python library called direct-redis. More on setting up MemoryStore later. 

MemoryStore provides a fully managed and low-cost platform for hosting Redis instances. Redis databases are in memory and light-weight making them an ideal candidate for caching. In the code above, we are caching the images fetched from Pixabay and their URLs. So, in the event of a cache hit, we won’t need to call the CLIP model and this will tremendously improve the response time of our API. 

Other options for caching

We can cache other elements of our application. For example, the natural language query. When searching through the cached entries to determine if it’s a cache hit, we can compare two queries for semantic similarity and return results accordingly. 

Consider that a user had entered the following natural language query: “mountains with dark skies”. After performing the search, we’d cache the embeddings of this query. Now, consider that another user entered another query: “mountains with gloomy ambiance”. We’d compute its embeddings and run a similarity search with the cached embeddings. We’d then compare the similarity scores with respect to a threshold and parse the most similar queries and their corresponding results. In case of a cache miss, we’d just call the image search utilities we developed above. 

When working on real-time applications we often need to consider these different aspects and decide what enhances the user experience and maximizes business at the same time. 

All that’s left now for the backend is our Flask application – main.py:

  @app.route("/search", methods=["GET"])
def get_images():
    tag = request.args.get("t").lower()
    query = request.args.get("s_query").lower()
    top_k = request.args.get("k")

    (top_urls, top_scores) = searcher.get_similar_images(
        tag, query, MAX_PIXABAY_SEARCH, int(top_k)
    )

    return jsonify({"top_urls": top_urls, "top_scores": top_scores})

Here we are first parsing the query parameters from the request payload of our search API.  We are then just calling the appropriate function from perform_search.py to handle the request. This Flask application is also capable of handling CORS. We do this via the flask_cors library:

  cors = CORS(
    app,
    resources={
        r"/search/*": {"origin": "*"},
        r"/test/*": {"origin": "*"},
    },
)

And this is it! Our API is now ready for deployment. 

Deployment with Compute Engine and GKE

The reason why we wanted to deploy our API on Kubernetes is because of the flexibility Kubernetes offers for managing deployments. When operating at scale, auto scalability and load balancing are very important. With the comes the requirement of security — we’d not want to expose the utilities for interacting with any internal services such as databases. With Kubernetes, we can achieve all these easily and efficiently. 

GKE provides secured and fully managed functionalities for operationalizing Kubernetes clusters. Here are the steps to deploy the API on GKE at a glance:

• We first build a Docker image for our API and then push it to the Google Container Registry (GCR).
• We then create a Kubernetes cluster on GKE and initialize a deployment.
• We then add scalability options.
• If any public exposure is needed for the API, we then tackle it. 

We can assimilate all the above into a shell script – k8s_deploy.sh:

  ## Docker build and push ## 
# We are inside the `server` directory
docker build -t gcr.io/${PROJECT_ID}/search_service .
docker push gcr.io/${PROJECT_ID}/search_service

## Deploy on a GKE cluster ##
# Configure the Docker command-line tool to authenticate to Container Registry
gcloud auth configure-docker

# Create a Kubernetes cluster
gcloud container clusters create image-search-nlp 

# Create a deployment
kubectl create deployment clip-search --image=gcr.io/${PROJECT_ID}/search_service

# Number of worker replicas
kubectl scale deployment clip-search --replicas=3

# HorizontalPodAutoscaler resource
kubectl autoscale deployment clip-search --cpu-percent=80 --min=1 --max=5

# Expose deployment
kubectl  expose deployment clip-search --name=clip-search-service --type=LoadBalancer --port 80 --target-port 8080

These steps are well explained in this tutorial that you might want to refer to for more details. We can configure all the dependencies on our local machine and execute the shell script above. We can also use the GCP Console to execute it since a terminal on the GCP Console is pre-configured with the system-level dependencies we’d need. In reality, the Kubernetes cluster should only be created once and different deployment versions should be created under it. 

After the above shell script is run successfully, we can run kubectl get service to know the external IP address of the service we just deployed:

  NAME                       CLUSTER-IP    EXTERNAL-IP     PORT(S)          AGE
clip-search-service      10.3.251.122    203.0.113.0     80:30877/TCP     10s

We can now consume this API with the following base URI: http://203.0.113.0/. If we wanted to deal with only http-based API requests, then we are done here. But secured communication is often a requirement in order for applications to operate reliably. In the following section, we are to discuss how to configure the additional items to allow our Kubernetes cluster to allow https requests.  

Configurations for handling https requests with GKE

A secure connection is almost often a  must-have requirement in modern client/server applications. The front-end Flutter application would be hosted on GitHub Pages for this project, and it requires https-based connection as well. Even if configuring https connection particularly for a GKE-based cluster can be considered a chore, its setup might seem daunting at first.

There are six steps to configure https connection in the GKE environment: 

1. You need to have a domain name, and there are a lot of inexpensive options that you can buy. For instance, mlgde.com domain for this project is acquired via Gabia which is a Korean service provider.
2. A reserved (static) external IP address has to be acquired via gcloud command or GCP console. 
3. You need to bind the domain name with the acquired external IP address. This is a platform-specific configuration that issued the domain name to you. 
4. There is a special ManagedCertificate resource which is specific to the GKE environment. ManagedCertificate resource specifies the domain that the SSL certificate will be created for, so you need this. 
5. An Ingress resource should be created by listing the static external IP address, ManagedCertificate resource, and the service name and port which the incoming traffic will be routed to. The Service resource could remain the same as in the above section with only changes from LoadBalancer to ClusterIP. 
6. Last but not least, you need to modify the existing Flask application and Deployment resource to support liveness and readiness probes which are used to check the health status of the Deployment. The Flask application side can be simply modified with the flask-healthz Python package, and you only need to add livenessProbe and readinessProbe sections in the Deployment resource. In the code example below, the livenessProbe and readinessProbe are checked via /alive and /ready endpoints respectively.

  from flask_healthz import healthz
from flask_healthz import HealthError


app = Flask(__name__)
app.register_blueprint(healthz, url_prefix="/")


def printok():
   print("Everything is fine")


def liveness():
   try:
       printok()
   except Exception:
       raise HealthError("Can't connect to the file")


def readiness():
   try:
       printok()
   except Exception:
       raise HealthError("Can't connect to the file")


app.config.update(
   HEALTHZ = {
       "alive": "main.liveness",
       "ready": "main.readiness",
   }
)

One thing to be careful of is the initialDelaySeconds attribute of the probes. It is uncommon to configure this attribute with a big number, but it could be bigger than 90 – 120 seconds depending on the size of the model to be used. For this project, it is configured in 90 seconds in order to wait until the CLIP model is fully loaded into memory (full YAML script here).

  livenessProbe:
 httpGet:
   path: /alive
   port: 8080
 initialDelaySeconds: 90
 periodSeconds: 10

readinessProbe:
 httpGet:
   path: /ready
   port: 8080
 initialDelaySeconds: 90
 periodSeconds: 10

Again, these steps may seem daunting at first, but it will become clear when you have done it once. Here is the official document for Using Google-managed SSL certificates You can find all the GKE-related resources used in this project here. 

Once every step is completed you should be able to see your server application running on the GKE environment. Please make sure to run kubectl apply command whenever you create Kubernetes resources such as Deployment, Service, Ingress, and ManagedCertificate, and it is important to wait for more than 10 minutes until the ManagedCertifcate provisioning is done. 

You can run gcloud compute addresses list command to find out the static external IP address that you have configured.

  NAME  ADDRESS/RANGE  TYPE      PURPOSE  NETWORK  REGION  SUBNET  STATUS
gde      34.149.231.34       EXTERNAL                                                                    IN_USE

Then, the IP address has to be mapped to the domain. Figure 3 is a screenshot of a dashboard from where we got the mlgde.com domain. It clearly shows mlgde.com is mapped to the static external IP address configured in GCP.

In case you’re wondering why we didn’t deploy this application on App Engine, well that is because of the compute needed to execute the CLIP model. App Engine instance won’t fit in that regime. We could have also incorporated compute-heavy capabilities via a VPC Connector. That is a design choice that you and your team would need to consider. In our experiments, we found the GKE deployment to be easier and suitable for our needs. 

Infrastructure for the CLIP model

As mentioned earlier, at the core of our application is the CLIP model. It is computationally a bit more expensive than the regular deep learning models. This is why it makes sense to have the hardware infrastructure set up accordingly to execute it. We ran a small benchmark in order to see how a GPU-based environment could be beneficial here. 

We ran the CLIP on a Tesla P100-based machine and also on a standard CPU-only machine 1000 times. The code snippet below is the meat of what we executed:

  DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

start_time = time.time()
for _ in range(1000):
    with torch.no_grad():
        model = model.to(DEVICE)
        inputs = processor(text=[semantic_search_phrase], 
                        images=all_images, return_tensors="pt", padding=True)
        inputs = inputs.to(DEVICE)
        outputs = model(**inputs)
end_time = time.time() - start_time
print(f"Total time: {end_time:.3f} seconds.")

As somewhat expected, with the GPU, the code took 13 minutes to complete execution. With no GPU, it took about 157 minutes.

It is uncommon to leverage GPUs for model prediction because of cost restrictions, but sometimes we have to access GPUs for deploying a big model like CLIP. We configured a GPU-based cluster on GKE and compared the performance differences with and without it. It took about 1 second to handle a request with GPU and MemoryStore cache while it took more than 4 seconds with MemoryStore only (without the GPUs). 

For the purposes of this post, we used a CPU-based cluster on Kubernetes. But It is easy to configure GPU usage in a GKE cluster. This document shows you how to do so. For a short summary, there are two steps. First, a node should be configured with GPUs when creating a GKE cluster. Second, GPU drivers should be installed in GKE nodes. You don’t need to visit and manually install GPU drivers for each node by yourself. Rather you can simply apply the DaemonSet resource to GKE as described here.

Setting up MemoryStore

In this project, we first query the general concept of images to Pixabay, then we filter the images with a semantic query using CLIP. It means we can cache the initially retrieved images from Pixabay for the next specific semantic query. For instance, you may want to search with “gentleman wearing tie” at first, then you may want to retry searching for “gentleman wearing glass”. In this case, the base images remain all the same, so they could be stored in a cache server like Redis. 

MemoryStore is a GCP service wrapping the Redis which is an in-memory data store, so you can simply use a standard Redis Python package for accessing it. The only thing to be careful about when provisioning a MemoryStore Redis instance is to make sure it is in the same region where your GKE cluster or Compute Engine instance is.

The code snippet below shows how to make a connection to the Redis instance in Python. Nothing specific to GCP, but you only need to be aware of the usage of the standard redis-py package. 

  # REDISHOST is the IP address to the MemoryStore instance
redis_host = os.environ.get("REDISHOST", "localhost")
redis_port = int(os.environ.get("REDISPORT", 6379))
redis_client = redis.StrictRedis(host=redis_host, port=redis_port)

After creating a connection, you can store and retrieve data from MemoryStore. There are more advanced use cases of Redis, but we only used exists, get, and set methods for the demonstration purpose. These methods should be very familiar if you know maps, dictionaries, or other similar data structures. For the code portion that uses Redis-related utilities, please refer to the Searcher Python class we discussed in an earlier section. 

In the URLs below, you can find side-by-side comparisons of using MemoryStore:

• Without MemoryStore: https://youtu.be/7B88Eyrd-4s
• With MemoryStore (1st try): https://youtu.be/LE6xeEIRuMM
• With MemoryStore (2nd try): https://youtu.be/rRfK17sdk84 

Putting everything together

All that’s left now is to collate the different components we developed in the sections above and deploy our application with a frontend. All the frontend-related code is present here. 

• The front-end application is written in the Flutter development kit. The main screen contains two text fields for queries to Pixabay and CLIP model respectively. When you click the “Send Query” button, it will send out a RestAPI request to the server. After receiving the result back from the server, the retrieved images from the semantic query will be displayed at the bottom section of the screen. 
  • Please note that a Flutter application can be deployed to various environments including desktop, web, iOS, and Android. In order to keep as simple as possible, we chose to deploy the application to the GitHub Pages. Whenever there is any change to a client-side source directory, the GitHub Action will be triggered to build a web page and deploy the latest version to the GitHub Pages. 

Our final application is deployed here and it looks like so:

Note that due to constraints, the above-mentioned URL will only be live for one or two months.  

It is also possible to redeploy the back-end application with a GitHub Action. 

• The very first step is to craft a Dockerfile like below. Since Python is a scripting language, and there are lots of heavy packages that the application is dependent on, it is important to cache the steps. For instance, installing the dependencies should be separated from other commands.

  FROM pytorch/pytorch:latest
WORKDIR /app

# install the dependencies
COPY requirements.txt requirements.txt
RUN pip install -r requirements.txt

COPY . .  

# set up environment variables
ENV PIXABAY_API_KEY="..."
ENV REDIS_IN_USE="true"
ENV REDISHOST="..."

# expose port that Flask app is listening on
EXPOSE 8080

# run the Flask app
CMD [ "python3", "main.py" ]

• With the Dockerfile defined, we can use a GitHub Action like this for automatic deployment. 

Edge cases

Since the CLIP model is pre-trained on a large corpus of image and text pairs it’s likely that it may not generalize well to every natural language query we throw at it. Also, because we are limiting the number of images on which the CLIP model can operate, this somehow restricts the expressivity of the model.

We may be able to improve the performance for the second situation by increasing the number of images to be pre-fetched and by indexing them into a low-cost and high-performance database like Datastore. 

Costs

In this section, we wanted to provide the readers a breakdown of the costs they might incur in order to consume the various services used throughout the application. 

• Frontend hosting
  • The front-end application is hosted on GitHub Pages, so there is no expenditure for this.
• Compute Engine
  • With an e2-standard-2 instance type without GPUs, the cost is around $48.92 per month. In case you want to add a GPU (NVIDIA K80), the cost goes up to $229.95 per month.
• MemoryStore
  • The cost for MemoryStore depends on the size. With 1GB of space, the cost is around $35.77 per month, and whenever you add more GBs the cost will be doubled.
• Google Kubernetes Engine
  • The monthly cost for a 3 node GKE cluster with n2-standard-2 (vCPUs: 2, RAM: 8GB without GPUs) is about $170.19. If you add one GPU (NVIDIA K80) to the cluster, the cost goes up to $835.48.

While you may think that is a lot cost-wise, it is good to know that Google gives away free $300 credits when you create a new GCP account. It is still not enough for leveraging GPUs, but it is enough to learn and experiment with GKE and MemoryStore usage.

Conclusion

In this post, we walked through the components needed to build a basic image search utility for natural language queries. We discussed how these different components are connected to each other. Our image search API is able to utilize caching and was deployed on a Kubernetes cluster using GKE. These elements are essential when building a similar service to cater to a much bigger workload. We hope this post will serve as a good starting point for that purpose. Below are some references on similar areas of work that you can explore:

• Building a real-time embeddings similarity matching system
• Detecting image similarity using Spark, LSH and TensorFlow

Acknowledgments: We are grateful to the Google Developers Experts program for supporting us with GCP credits. Thanks to Karl Weinmeister and Soonson Kwon of Google for reviewing the initial draft of this post. 

Editor’s note: Today we’re hearing from Ryan Mason, Director, Head of DTC Growth & Strategy, at alcoholic beverage firm, Constellation Brands on the company’s shift to Direct-to-Consumer (DTC) sales and how Google Cloud’s powerful technology stack helped with this transformation. 

It’s no secret that consumer businesses have been up-ended in a lasting manner after 18 months of the pandemic. Consumers have been forced to shop differently over the past year – and as a result, they’ve evolved to be more comfortable with online spending and have grown to expect a certain level of convenience. While the e-commerce share of consumer sales has grown steadily over the past decade, the pandemic was the catalyst for the famous “10 years of growth in 3 months” which many argue is here to stay. 

Facing this reality head-on,  we placed a new emphasis on Direct-to-Consumer (DTC) with our acquisition of Empathy Wines, a DTC-native wine brand that sells directly to consumers via e-commerce. To accelerate our innovation in the DTC space, we added headcount and new functions to the existing Empathy team and empowered the newly-minted DTC group to apply their digital commerce operating model across the rest of the wine and spirits portfolio, which includes Robert Mondavi Winery, Meiomi Wines, The Prisoner Wine Company, High West Whiskey, and more. 

One pandemic and one year later, DTC sales have surged in the wine and spirits category with Constellation positioned as a leader armed with a unique and powerful cloud technology stack, best-in-class e-commerce user experiences, modernized fulfillment solutions, and data-driven growth marketing. 

Benefits of Going DTC 

A report from McKinsey estimates that the strategic business shift to DTC has been accelerated by two years because of the pandemic and argues that consumer brands that want to thrive will need to aim for a 20% DTC business or higher, which is already taking shape in the market: Nike’s direct digital channels are on track to make up 21.5% of the total business by the end of 2021, up from 15.5% in the last fiscal year, and Adidas is aiming for 50% DTC by 2025. But outside of the clear revenue upside, the auxiliary benefits of going DTC are robust.

For Constellation Brands, each of these four pillars ring true, and our shift toward DTC is as much about margin accretion and revenue mix management as it is about consumer insights and data. The added complexities of the alcohol space add wrinkles to our DTC approach and manifest in many areas like consumer shopping preference, shipping and logistics hurdles, and more. In order to win share early and continue to lead the category, we recognized the need to harness the immense amount of first-party data to power impactful and actionable insights. 

Our DTC technology architecture has fostered a value chain that is completely digitized: website traffic, marketing expenditures, tasting room transactions, e-commerce transactions, logistics and fulfillment events, cost of goods sold (COGS) and margin profiles, etc. are recorded and stored in a data warehouse in real time.  For the first time, at any given moment, we can easily and deterministically answer complex business questions like “what is the age and gender distribution of my customers from Los Angeles who have purchased SKU X from Brand.com Y in the last 6 months? What is the cohort net promoter score? Did that increase after we introduced same-day shipping in this zip code? By how much?” 

The ability to answer these questions and understand the root causes allows us to stay nimble with product offerings and iterate marketing strategies at the speed of consumer preference. Further, it enables us to optimize our omnichannel presence in the same manner by leaning on DTC consumer insights to develop valuable strategies with key wholesale distribution partners and 3-Tier eCommerce partners like Drizly and Instacart. At its core, Constellation’s DTC practice is designed to be the consumer-centric “tip-of-the-spear” responsible for generating insights from which all sales channels, including wholesale, can benefit. 

Constellation’s DTC technology approach prioritizes consumer-centricity and insights generation

We have taken a modern approach to building a digital commerce technology stack, leveraging a hub-and-spoke model built around Shopify Plus and other key emergent technology providers like email provider Klaviyo, loyalty platform Yotpo, Net Promoter Score measurer Delighted, Customer Service module Gorgias, payments processor Stripe, event reservations platform Tock, and many more. For digital marketing and analytics, we use Google Cloud and Google Marketing Platform, which includes products like Analytics 360, Tag Manager 360, and Search Ads 360.

To help gather, organize, and store all of the inbound data from the ecosystem, we partnered with SoundCommerce, a data processing platform for eCommerce businesses. Together with SoundCommerce, we are able to automate data ingestion from all endpoints into a central data warehouse in Google BigQuery. With BigQuery, our data team is able to break data silos and quickly analyze large volumes of data that help unlock actionable insights about our business.  BigQuery itself allows for out-of-the-box predictive analytics using SQL via BigQuery ML, and a key differentiator for us is that all Google Marketing Platform data is natively accessible for analysis within BigQuery. 

But data possession only addresses half of the opportunity: we needed a powerful and modern business intelligence platform to help make sense of the vast amounts of data flowing into the system. Core to the search was to find a partner that approached BI in a way that fit with our future-looking strategy.

Our DTC team relies on the accurate measurement of variable metrics like Customer Acquisition Cost (CAC), Customer Lifetime Value (CLV), Churn, and Net Promoter Score (NPS) as a bellwether of the health of the business and monitoring these figures on a daily basis is paramount to success. To enable us to keep an accurate pulse on strategic KPIs, we considered several incumbent BI platforms. Ultimately we selected Google Cloud’s Looker for a range of benefits that separated it from the rest of the pack.

From a vision perspective, in this particular case we felt Looker was most aligned with our belief that better decisions are made when everyone has access to accurate, up-to-date information. Looker allows us to realize that vision by surfacing data in a simple web-based interface that empowers everyone to take action with real-time data on critical commercial activities. Furthermore, Looker’s ability to automate and distribute formatted modules to a myriad of stakeholders on a regular cadence increases data literacy and business performance transparency.

From a product perspective, we chose Looker for it’s cloud offering, web-based interface, and  centralized, agile modeling layer that creates a trusted environment for all users to confidently interact with data — without any actual data extraction. While other BI tools have centralized semantic layers that require skilled IT resources, we’ve experienced that those can lead to bottlenecks and limited agility. With Looker’s semantic layer, LookML, our BI Team, led by Peter Donald, can easily build upon their SQL knowledge to add both a high degree of control as well as flexibility to our data model.  The fully browser-based development environment allows the data team to rapidly develop, test, and deploy code and is backed by robust and seamless Git source code management. 

In parallel, LookML empowers  business users to collaborate without the need for advanced SQL knowledge. Our data team curates interactive data experiences with Looker to help scale access and adoption. Business users can explore ad hoc analysis, create dashboards, and develop custom data experiences in the web-based environment to get the answers they need without relying on IT resources each time they have a new question, while also maintaining the confidence that the underlying data will always be accurate. This helps us meet our primary goal of providing all businesses users with the data access they need to monitor the pulse of key metrics in near real-time.

Impact and future of DTC BI at Constellation

In short order, taking a modern and integrated approach to the DTC technology stack has delivered economic impact across the portfolio, helping our team understand and combat customer churn, increase conversion rates, and optimize the customer acquisition cost (CAC) and customer lifetime value (CLV) ratios. Perhaps most important is the benefit it can provide to the customer base. Mining customer data and consumer behavior generates data into what our customers are seeking, giving us insights to supply more, or less of it. For example, observing sales velocity and conversion rates by SKU or by region can help us better understand changes in customer taste profiles and fluctuations in demand, providing the foundation for a more powerful innovation pipeline and more effective sales and distribution tactics in wholesale. Our team has also been an early pilot tester for Looker’s new integration with Customer Match, which contributes to the virtuous cycle between data insight and data activation. In the future, our plan is to leverage this cycle to amplify the impact of Google Ads across Search, Shopping, and YouTube placements for the wine and spirits portfolio. 

The operational impact of Looker is also substantial: our team estimates that the number of hours needed to reach critical business decisions has been reduced by nearly 60%, boosting productivity and accelerating the daily operating rhythm. A thoughtfully curated technology stack together with a modern BI solution allows us to stay at the vanguard of the industry. While the DTC sales channel is not designed to surpass the core business of wholesale for Constellation in terms of size, the approach enables unparalleled insights and measurement abilities that will pay dividends for the entire business for years to come.