As software-defined vehicles continue to advance and the quantity of digital services grows to meet consumer demand, the data required to provide these services continue to grow as well. This makes automotive manufacturers and suppliers look for capabilities to log and analyze data, update applications, and extend commands to in-vehicle software.

The challenges the automotive sector faces can be quantified. A modern vehicle contains upwards of 70 electronic control units (ECUs), most of which are connected to one or more sensors. Not only is it now possible to exactly measure many aspects of vehicle performance, but new options become available. Using LIDAR (light detection and ranging), for example, vehicles are achieving higher levels of autonomy; this leads to a data stream from such demanding applications that may reach 25 GB per hour. For the in-vehicle processing of data, 100 million lines of software code may be present — more than a fighter jet. This in-vehicle code will have to be maintained with updates and new functionalities.

Access to the data will allow manufacturers to gain valuable insights into operational details of their vehicles. The use of this data can help to reduce costs and risks, increase ROI, support ESG initiatives, and provide valuable insights to develop innovative solutions and shorten the time to value for Electric Vehicle innovations.

Sibros’ Deep Connected Platform (DCP) makes it possible for these manufacturers to build and launch new connected vehicle use cases from production to post-sale at scale by connecting and managing all software and data throughout every life cycle stage. A key component of this platform is the Sibros Deep Logger that provides capabilities like the following:

• Full configurability of what to record, when to record it, and how fast to record it.
• High resolution timestamps of all Controller Area Network (CAN) messages.
• Dynamic application of live log configurations to receive new data points without deploying new software.

For example, properly analyzed engine data enables true predictive maintenance for the first time, which creates the option to repair or replace components before failure happens. Another example would be the evaluation of data regarding the use of certain in-car features with the goal to redesign its interior.

Two other components of the DCP are software updates and remote commands to ECUs. The DCP on Google Cloud enables seamless integration with any vehicle architecture and provides OEMs and suppliers with the platform to manage connected vehicle data at rest and in transit using a proven and secure way on a global scale.

OEMs can pull data through APIs provided by Sibros into Google Data Cloud (including BigQuery) to gain access to the rich information data sets provided by the DCP within their environment and blend this data with their first party data sets to provide value insights for their business. Some of the Connected Vehicle insights that DCP information enables are:

• Damage prevention, improved operation, or development of the next generation of engines with insights from complex analyses that could consider parameters like model, engine type, mileage, overall speed, temperature, air pressure, load, services, and more.
• The combination of electric vehicle battery usage data like charging cycles, engine performance, and battery age with contributing factors as the use of the air conditioning to determine if such factors contribute to hazardous battery conditions and for improved battery development.
• Cross-organization collaboration in R&D by the provision of information on all these metrics and more from real-world driving, like engine knock data and even tire pressure.
• Google Cloud’s unified data cloud offering provides a complete platform for building data-driven applications like those from Sibros — from simplified data ingestion, processing, and storage to powerful analytics, AI, ML, and data sharing capabilities — integrated with Google Cloud. With a diverse partner ecosystem and support for multi-cloud, open-source tools and APIs, Google Cloud provides Sibros the portability and the extensibility they need to avoid data lock-in.

“Software has an ever increasing importance in the automotive world, even more so with electric vehicles and new mobility services. Google Cloud is partnering with Sibros to bring their award winning Deep Connected Platform to deliver high frequency, low latency over-the-air software updates, data logging & diagnostics capabilities to our automotive customers, leveraging the security and scale of Google Cloud. This is revolutionizing everything from development cycles to business models and customer relationships.” — Matthias Breunig, Director, Global Automotive Solutions, Google Cloud

Through Built with BigQuery, Google Cloud is helping tech companies like Sibros build innovative applications on Google’s Data Cloud with simplified access to technology, helpful and dedicated engineering support, and joint go-to-market programs.

“Sibros is looking forward to partnering with Google Cloud, which will enable vehicle manufacturers and suppliers to reach the next level in their use of data. Sibros solutions for Deep Data Logging and Updating on the Google Data Cloud, combined with Google BigQuery, will help them to mitigate risks, reduce costs, add innovative products, and introduce value-added use cases.” — Xiaojian Huang, Chief Digital Officer, Software, Sibros

Sibros and Google Cloud are driving Connected Mobility transformation to help our customers accelerate R&D innovation, power efficient operations, and unlock software-defined vehicle use cases with a full stack connected vehicle platform. Click here to learn more about Sibros on Google Cloud.

Over the past two decades, consumers have become accustomed to receiving personalized recommendations in all facets of their online life. Whether that be recommended products while shopping on Amazon, a curated list of apps in the Google Play store, or relevant videos to watch next on YouTube. In fact, in a Verge article “​​How YouTube perfected the feed: Google Brain gave YouTube new life,” the Google Brain team reveals how their recommendation engine has impacted the platform with “more than 70 percent of the time people spend watching videos on the site being driven by YouTube’s algorithmic recommendations” thereby increasing time spent on the platform by 20X in three years. 

It’s become clear that personalized recommendations are no longer a differentiator for an organization but rather something consumers have come to expect in their day-to-day experiences online. So what should you do if you are behind the curve and want to get started or simply want to improve upon what you already have? While there are all sorts of techniques, from content-based systems to deep learning methods, our goal in this recommender-focused blog series is to demystify three available approaches to building recommendation systems on Google Cloud: Matrix Factorization in BigQuery Machine Learning (BQML),  Recommendations AI, and deep retrieval techniques available via the Two-Tower built-in algorithm.

One of these approaches can be used to meet you where you are in your personalization journey, no matter if you are just starting or if you are well into it. This first blog post will introduce our three approaches and when to use them. 

What is Matrix Factorization and how does it work?

Collaborative filtering is a foundational model for building a recommendation system as the input dataset is simple and the embeddings are learned for you. How does Matrix factorization fit into the mix you might be wondering?  Matrix factorization is simply the model that applies collaborative filtering. BQML enables users to create and execute a matrix factorization model by using standard SQL directly in the data warehouse. 

Collaborative filtering begins by creating an interaction matrix. The interaction matrix represents users as a row and items as columns in your dataset. This interaction matrix often is sparse in nature as not all users will have interacted with many items in your catalog. This is where embeddings come into play. Generating embeddings for users and items not only allows you to collapse many sparse features into a lower dimensional space but they also allow you to derive a similarity measure so that similar users/items fall nearby in the embedding space. These similarity measures are key as collaborative filtering uses similarities between users and items to make the end recommendations. The underlying assumption being that similar users will like similar items whether that be movies or handbags. 

What’s required to get started?

To train a matrix factorization model you need a table that includes three input columns: user(s), item(s), and an implicit or explicit feedback variable (e.g., ratings is an example of explicit feedback). With the base input dataset in place, you can then easily run your model in BigQuery after specifying several hyperparameters in your CREATE MODEL SQL statement. Hyperparameters are available to specify the number of embeddings, the feedback type, the amount of L2 regularization applied and so on.    

Why use this approach and who is it a good fit for?  

As mentioned earlier, Matrix Factorization in BQML is a great way for those new to recommendation systems to get started. Matrix factorization has many benefits: 

• Little ML Expertise: Leveraging SQL to build the model lowers the level of ML expertise needed
• Few Input Features: Data inputs are straightforward, requiring a simple interaction matrix
• Additional Insight: Collaborative filtering is adept at discovering new interests or products for users 

While Matrix Factorization is a great tool for deriving recommendations it does come with additional considerations and potential drawbacks depending upon the use case. 

• Not Amenable to Large Feature Sets: The input table can only contain two feature columns (e.g., user(s), item(s)). If there is a need to include additional features such as contextual signals, Matrix factorization may not be the right method for you.  
• New Items: If an item is not available in the training data, the system can’t create an embedding for it and will have difficulty recommending similar items. While there are some workarounds available to address this cold-start issue, if your item catalog often includes new items, Matrix factorization may not be a good fit.  
• Input Data Limitations: While the input matrix is expected to be sparse, training examples without feedback can cause problems. Filtering for items and users that have at least a handful of feedback (e.g., ratings) examples can improve the model. More information on limitations can be found here. 

In summary, for users with a simplified dataset looking to iterate quickly and develop a baseline recommendation system, Matrix Factorization is a great approach to begin your personalization AI journey. 

What is Recommendations AI and how does it work?

Recommendations AI is a fully managed service which helps organizations deploy scalable recommendation systems that use state-of-the-art deep learning techniques, including cutting-edge architectures such as two-tower encoders, to serve personalized and contextually relevant recommendations throughout the customer journey.

Deep learning models are able to improve the context and relevance of recommendations in part because they can easily address the previously mentioned limitations of Matrix Factorization. They incorporate a wide set of user and item features, and by definition they emphasize learning successive layers of increasingly meaningful representations from these features. This flexibility and expressivity allows them to capture complex relationships like short-lived fashion trends and niche user behaviors. However, this increased relevance comes at a cost, as deep learning recommenders can be difficult to train and expensive to serve at scale. 

Recommendations AI helps organizations take advantage of serving these deep learning models and handles the MLOps required to serve these models globally with low latency. Models are automatically retrained daily and tuned quarterly to capture changes in customer behavior, product assortment, pricing, and promotions. Newly trained models follow a resilient CI/CD routine which validates they are fit to serve and promotes them to production without service interruption. The models achieve low serving latency by using a scalable approximate nearest neighbors (ANN) service for efficient item retrieval at inference time. And, to maintain consistency between online and offline tasks, a scalable feature store is used, preventing common production challenges such as data leakage and training-serving skew.  

What’s required to get started?

To get started with Recommendations AI we first need to ingest product and user data into the API:

• Import product catalog: For large product catalog updates, ingest catalog items in bulk using the catalogItems.import method. Frequent catalog updates can be schedule with Google Merchant Center or BigQuery
• Record user events: User events track actions such as clicking on a product, adding items to cart, or even purchasing an item. These events need to be ingested in real time to reflect the latest user behavior and then joined to items imported in the product catalog 
• Import historical user events: The models need sufficient training data before they can provide accurate predictions. The recommended user event data requirements are different across model types (learn more here)

Once the data requirements are met, we are able to create one or multiple models to serve recommendations: 

• Determine your recommendation types and placements:  The location of the recommendation panel and the objective for that panel impact model training and tuning. Review the available recommendations types, optimization objectives, and other model tuning options to determine the best options for your business objectives.
• Create model(s): Initial model training and tuning can take 2-5 days depending on the number of user events and size of the product catalog 
• Create serving configurations and preview recommendations: After the model is activated, create serving configurations and preview the recommendations to ensure your setup is functioning as expected before serving to production traffic

Once models are ready to serve, consider setting up A/B experiments to understand how newly trained models impact your customer experience before serving them to 100% of your traffic. In the Recommendations AI console, see the Monitoring & Analytics  page for summary and placement-specific metrics (e.g., recommender-engaged revenue, click-through-rate, conversion rate, and more).

Why use this approach and who is it a good fit for?  

Recommendations AI is a great way to engage customers and grow your online presence through personalization. It’s used by teams who lack technical experience with production recommendation systems, as well as customers who have this technical depth but want to allocate their team’s effort towards other priorities and challenges. No matter your team’s technical experience or bandwidth, you can expect several benefits with Recommendations AI: 

• Fully managed service: no need to preprocess data, train or hypertune machine learning models, load balance or manually provision you infrastructure – this is all taken care of for you. The recommendation API also provides a user-friendly console to monitor performance over time. 
• State-of-the-art AI: take advantage of the same modeling techniques used to serve recommendations across Google Ads, Google Search, and YouTube. These models excel in scenarios with long-tail products and cold-starts users and items
• Deliver at any touchpoint: serve high-quality recommendations to both first-time users and loyal customers anywhere in their journey via web, mobile, email, and more
• Deliver globally: serve recommendations in any language anywhere in the world at low-latency with a fully automated global serving infrastructure
• Your data, your models: Your data and models are yours. They’ll never be used for any other Google product nor shown to any other Google customer

For users looking to leverage state of the art AI to fuel their recommendation systems but need an existing solution to get up and running more quickly, Recommendations AI is the right solution for you. 

What are Two Tower encoders and how do they work?

As a reminder, in recommendation system design, our objective is to surface the most relevant set of items for a given user or set of users. The items are usually referred to as the candidate(s) where we might include information about the items such as the title or description of the item, other metadata about the item like language, number of views, or even clicks on the item over time. User(s) are often represented in the form of a query to a recommendation system where we might provide details about the user such as the location of the user, preferred languages, and what they have searched for in the past.   

Let’s start with a common example. Imagine that you are creating a movie recommendation system. The input candidates for such a system would be thousands of movies and the query set can consist of millions of viewers. The goal of the retrieval stage is to select a smaller subset of movies(candidates) for each user and then score and rank order them before presenting the final recommended list to the query/user.

The retrieval stage is able to refine our list of candidates by encoding both the candidate and the query data so they share the same embedding space. A good embedding space will place candidates which are similar to one another closer together and dissimilar items/queries farther apart in the embedding space.

Once we have a database of query and candidate embeddings we can then use an approximate nearest neighbor search method to then generate a list of final “like” candidates, i.e. find a certain number of nearest neighbors for a given query/user and surface final recommendations.

What’s required to get started?

At the most basic level, in order to train a two-tower model you need the following inputs:

• Training Data: Training data is created by combining your query/user data with data about the candidates/items. The data must include matched pairs, cases where both user and item information is available. Data in the training set can include many formats from text, numeric data, or even images. 
• Input Schema: The input schema describes the schema of the combined training data along with any specific feature configurations.

Several services within Vertex AI have come available that complement the existing Two-Tower built-in algorithm and can be leveraged in your execution: 

• Nearest Neighbor (ANN) Service: Vertex AI Matching Engine and  ScANN provide a high-scale and low-latency Approximate Nearest Neighbor (ANN) service so you can more easily identify similar embeddings.
• Hyperparameter Tuning Service: A hyperparameter tuning service such as Vizier can help you identify the optimal hyperparameters such as the number of hidden layers, the size of the hidden layers, and the learning rate in fewer trials. 
• Hardware Accelerators: Specialized hardware, such as GPUs or TPUs, can be valuable in your recommendation system to help accelerate experiments and improve the speed of training cycles. 

Why use this approach and who is it a good fit for?  

The Two-Tower built-in algorithm can be considered the “custom sports car” of recommendation systems and comes with several benefits: 

• Greater Control: While Recommendations AI uses the two-tower architecture as one of the available architectures it doesn’t provide granular control or visibility into model training, example generation, and model validation details. In comparison, the Two-Tower built in algorithm provides a more customizable approach as you are training a model directly in a notebook environment. 

• More Feature Options: The Two Tower approach can handle additional contextual signals ranging from text to images. 
• Cold Start Cases: Leveraging a rich set of features not only enhances performance but also allows the candidate generation to work for new users or new candidates.

While the Two-Tower built in algorithm is an excellent and best-in class solution for deriving   recommendations, it does come with additional considerations and potential drawbacks depending upon the use case.

• Technical ML Expertise Required: Two tower encoders are not a “plug and play” solution like the other approaches mentioned above. In order to effectively leverage this approach, appropriate coding and ML expertise is required. 
• Speed to Insight: Building out a custom solution via two-tower encoders may require additional time as the solution is not pre-built for the user.   

For users looking for greater control, increased flexibility, and have the technical chops to easily work within a managed notebook environment – the two-tower built in algorithm is the right solution for them. 

What’s next?

In this article, we explored three common methods for building recommendation systems on Google Cloud Platform. As you can see thus far, there are alot of considerations to take into account before choosing a final approach. In an effort to help you align more quickly we have distilled the decision criteria down to a few simple steps (see below for more details).

In the next installments of this series, we will dive more deeply into each method, explore how hardware accelerators can play a key role in recommendation system design, and discuss how recommendation systems may be leveraged in key verticals. Stay tuned for future posts in our recommendation systems series. Thank you for reading! Have a question or want to chat? Find authors here – R.E. [Twitter | LinkedIn], Jordan [LinkedIn], and Vaibhav [LinkedIn].

Acknowledgements

Special thanks to Pallav Mehta, Henry Tappen,Abhinav Khushraj, and Nicholas Edelman for helping to review this post. 

References

• Announcing ScaNN: Efficient Vector Similarity Search
• Factorization Machines for Item Recommendation with Implicit Feedback Data

Your product teams might ask – “Why does it take so long to build a feature or application?” Building applications is a heavy lift due to the technical complexity, which includes the complexity of backend services that are used to manage and store data. Every moment focused on this technical complexity is a distraction from  delivering on core business value. Firestore alters this by having Google Cloud manage your backend complexity through a complete backend-as-a-service! 

Firestore is a serverless, NoSQL document database that unlocks application innovation with simplicity, speed and confidence. 

It acts as a glue that intelligently brings together the complete Google Cloud backend ecosystem, in-app services from Firebase and core UI frameworks & OS from Google.

What is Firestore?

Firestore is a serverless, fully managed NoSQL document database that scales from zero to global scale without configuration or downtime. Here’s what makes Firestore unique:

• Ideal for rapid, flexible and scalable web and mobile development with direct connectivity to the database.
• Supports effortless real time data synchronization with changes in your database as they happen. 
• Robust support for offline mode, so your users can keep interacting with your app even when the internet isn’t available or is unreliable.
• Fully customizable security and data validation rules to ensure the data is always protected
• Built-in strong consistency, elastic scaling, high performance & best in class 99.999% availability 
• Integration with Firebase and Google Cloud services like Cloud Functions and BigQuery, serverless data warehouse.
• In addition to a rich set of Google Cloud service integrations, Firestore also offers deep one-click integrations with a growing set of 3rd party partners via Firebase Extensions to help you even more rapidly build applications.

Document-model database

Firestore is a Document-model database. All of your data is stored in “documents” and then “collections”.  You can think of a document as a JSON object. It’s a dictionary with a set of key-value mappings, where the values can be several different supported data types including strings, numbers or binary values..

These documents are stored in collections. Documents can’t directly contain other documents, but they can point to subcollections that contain other documents, which can point to subcollections, and so on. This structure brings with it a number of advantages. For starters, all queries that you make are shallow, meaning that you can grab a document without worrying about grabbing all the data underneath it. And this means that you can structure your data hierarchically in a way that makes sense to you logically, without having to worry about grabbing tons of unnecessary data. 

How to use Firestore?

Firestore can be used in two modes:

• Firestore in Native Mode:  This mode is differentiated by its ability to directly connect your web & mobile app to Firestore. Native Mode supports up to 10K writes per second, and over a million connections. 
• Firestore in Datastore Mode: This mode supports only server-side usage of Firestore, but supports unlimited scaling, including writes. 

Conclusion

Whatever your application use case may be, if you want to build a feature or an application quickly using Firestore backend-as-a-service. For a more in-depth look into Firestore check out the documentation. https://www.youtube.com/embed/moglAjmwmUQ?enablejsapi=1&

For more #GCPSketchnote, follow the GitHub repo. For similar cloud content follow me on Twitter @pvergadia and keep an eye out on thecloudgirl.dev.

This is a guest post authored by Firebase customer, Anton Ivanov, Founder & CEO of DealCheck

Real estate investing is a fantastic way to build a stream of passive income and grow your wealth. Numerous studies have pointed out that real estate investing has created more millionaires throughout history than any other form of investing (like this one and this one). So why don’t more people do it?

I asked myself this very question a few years ago after talking to a group of friends about the success I’ve had with real estate, and listening to their reasons why they think it’s out of their reach.

A common theme among them was that they viewed it as something too difficult to learn and master. There were too many steps, the learning curve was steep and there was a lot of room for mistakes for somebody just starting out, especially when analyzing the financial performance of potential investment properties.

Traditionally, most investors used spreadsheets to do the math – which works only if you know what and how you’re calculating something. But if you don’t know that, it’s very easy to make mistakes and overlook things. And no one wants to make mathematical errors before a huge purchase like an investment property.

Where do I even begin?!

And that’s when I had the idea to build DealCheck – a cloud-based, easy-to-use property analysis tool for real estate investors and agents. I wanted to create a platform that would help new investors learn the ropes and avoid costly mistakes, but at the same time provide the flexibility to perform more advanced analysis with a click of a button.

Making real estate investing easier and more accessible.

The Challenges of Solo Development

I was working as a front-end engineer at the time, so I knew I could build the UI myself, but what about the back-end, data storage, authentication, and a bunch of other things you need for a full-functioning cloud app?

I didn’t know anybody I could bring on as a co-founder, so I set out to research what technologies and platforms I could leverage to help me with the back-end and server infrastructure.

Firebase kept popping up again and again and I began to look at it in more detail. It was then recently acquired by Google and its collection of BaaS (backend-as-a-service) modules seemed to offer the exact solution I needed to build DealCheck.

I was especially impressed with the documentation for each feature and how well all of the different technologies could be tied together to create one unified platform.

It wasn’t long before I signed up and started building the first MVP of the app.

Using Firebase to Quickly Build a Scalable Backend

As the only developer on the project, I had limited time and resources to spend on building the back-end, so I set out to use every Firebase feature that was available at the time to my advantage.

My goal was actually to write as little server-side code as possible and instead focus on leveraging the different Firebase modules to solve three specific challenges:

Challenge #1 – Authentication and User Management

The first one was authentication and user management. DealCheck’s users needed the ability to create their accounts so they can view and analyze properties on any device (more on that later). I wanted to have the ability to sign in with email, Facebook or a Google account.

Firebase Authentication was designed specifically for this purpose and I used it to handle pretty much the entire authentication flow. Out-of-the-box, it has support for all the major social networks, cross-network credential linking and the basic account management operations like email changes, password resets and account deletions.

There was no server-side code required at all – I just needed to build the UI on the front-end.

Email, Facebook and Google sign in powered by Firebase.

And as an added benefit, Firebase Authentication ties directly into the Realtime Database product to create a declarative permissions and access control framework that’s easy to implement and maintain. This helped me make sure user data was protected from unauthorized access, but also facilitate data sharing among users.

Challenge #2 – Cloud Storage with Cross-Device Sync

Next up was data storage. I knew that I wanted DealCheck’s users to be able to use the app and analyze properties online, on iOS and Android. So I needed a real-time, cloud-based database solution that could sync data across any device.

Syncing data across web and mobile is not easy!

Firebase Realtime Database is a NoSQL, JSON-based database solution that was designed exactly for this purpose, and I was actually surprised how great it worked. I used the official AngularJS bindings for Firebase on the front-end to read and write to it directly from the client.

I had to do some extra work on mobile to implement an offline mode with syncing after reconnections, but all-together the code required to make everything work was minimal.

As I mentioned, Firebase Authentication tied directly to the database to facilitate access control, so I really didn’t need to do anything extra there. And I was able to set up automatic daily backups of all the data with a click of a button.

Challenge #3 – Third-Party Integrations

Up to now, I had written exactly 0 lines of server-side code and everything was handled by the client directly. As DealCheck’s development progressed, however, I knew that I would need a server to handle some operations that could not be done in the client.

I wasn’t very experienced with server maintenance and DevOps, but fortunately the Firebase Cloud Functions product was able to solve all of my needs. Cloud Functions are essentially single-purpose functions that can be triggered (or executed) based on a specific HTTP request or events coming from the Authentication, Realtime Database or other Firebase products.

Each function can be run once based on a specific event trigger to perform its prescribed task. You don’t have to worry about provisioning a server instance or managing load – everything is done automatically for you by Firebase.

What’s even cooler, is that Cloud Functions can access the Realtime Database and Cloud Storage buckets of the same project, performing operations on them server-side, as needed.

This is how DealCheck processes subscription payments through Stripe, validates Apple and Google Play mobile subscription receipts, integrates with third-party APIs and updates database records without user interaction.

Bringing in sales comparable data from third-party providers into DealCheck.

Cloud Functions became the “glue” that tied the entire back-end infrastructure together.

Growing from an MVP to 100,000 Users with Firebase

The first version of the DealCheck app was built and launched in less than 5 months with just me on the development team. I definitely don’t think that would have been possible without Firebase powering the back-end infrastructure. Maybe the project wouldn’t have ever launched at all.

While Firebase is awesome for quick MVP development, it’s definitely designed to power production applications at scale as well. As DealCheck grew from a small side-project to one of the most popular real estate apps with over 100k users, all of the Firebase products that we use scaled to support the increasing load.

Moreover, the fantastic interoperability of all Firebase modules allows us to develop and release new features much faster because of the reduced coding requirements and ease of configuration.

So next time you’re looking to build an ambitious project with a small team – take a look at how Firebase can help you reduce development time and provide a suite of powerful tools that scale as your business grows.

This is exactly how DealCheck grew from a simple idea to make property analysis easier and faster, to an app that is helping tens of thousands of people grow their wealth and passive income through real estate investing. It’s a truly awesome and fulfilling experience to see your work positively impact so many people and it wouldn’t have been possible without Firebase.