Digitec Galaxus AG is the biggest online retailer in Switzerland, operating two online stores: Digitec, Switzerland’s online market leader for consumer electronics and media products, and Galaxus, the largest Swiss online shop with a steadily growing range of consistently low-priced products for almost all daily needs. 

Known for its efficient, personalized shopping experiences, it’s clear that Digitec Galaxus understands what it takes to deliver a platform that is interesting and relevant to customers every time they shop. 

The problem: Personalizing decisions for every situation

Digitec Galaxus already had established an engine to help them personalize experiences for shoppers when they reached out to Google Cloud. They had multiple recommendation systems in place and were also extensive early adopters of Recommendations AI, which already enabled them to offer personalized content in places like their homepages, product detail pages, and their newsletter. 

But those same systems sometimes made it difficult to understand how best to combine and optimize to create the most personalized experiences for their shoppers. Their requirements were threefold:

1. Personalization: They have over 12 recommenders they can display on the app, however they would like to contextualize this and choose different recommenders (which in turn select the items) for different users. Furthermore they would like to exploit existing trends as well as experiment with new ones.
2. Latency: They would like to ensure that the solution is architected so that the ranked list of recommenders can be retrieved with sub 50 ms latency.
3. End-to-end easy to maintain & generalizable/modular architecture: Digitec wanted the solution to be architected using an easy to maintain, open source stack, complete with all MLops capabilities required to train and use contextual bandits models. It was also important to them that it is built in a modular fashion such that it can be adapted easily to other use cases which have in mind such as recommendations on the homepage, Smartags and more . 

To improve, they asked us to help them implement a machine learning (ML) contextual bandit based recommender system on Google Cloud taking all the above factors into consideration to take their personalization to the next level. 

Contextual bandits algorithms are a simplified form of reinforcement learning and help aid real-world decision making by factoring in additional information about the visitor (context) to help learn what is most engaging for each individual. They also excel at exploiting trends which work well, as well as exploring new untested trends which can yield potentially even better results. For instance, imagine that you are personalizing a homepage image where you could show a comfy living room couch or pet supplies. 

Without a contextual bandit algorithm, one of these images would be shown to someone at random without considering information you may have observed about them during previous visits. Contextual bandits enable businesses to consider outside context, such as previously visited pages or other purchases, and then observe the final outcome (a click on the image) to help determine what works best. 

Creating a personalization system with contextual bandits

While Digitec Galaxus heavily personalizes their website homepages, they are very very sensitive and also require more cross-team collaboration to update and make changes. 

Together with the Digitec Galaxus team, we decided to narrow the scope and focus on building a contextual bandit personalization system for the newsletter first. The digitec Galaxus team has complete control over newsletter decisions and testing various ML experiments on a newsletter would have less chance of adverse revenue impact than a website homepage. 

The main goal was to architect a system that could be easily ported over to the homepage and other services offered by Digitec with minimal adaptations. It would also need to satisfy the functional and non-functional requirements of the homepage as well as other internal use cases.

Below is a diagram of how the newsletter’s personalization recommendation system works:

• The system is given some context features about the newsletter subscriber such as their purchase history and demographics. Features are sometimes referred to as variables or attributes, and can vary widely depending on what data is being analyzed. 
• The contextual bandit model trains recommendations using those context features and 12 available recommenders (potential actions). 
• The model then calculates which action is most likely to enhance the chance of reward (a user clicking in the newsletter) and also minimize the problem (an unsubscribe). 

Calculating whether a click was a newsletter or an unsubscribe enabled the system to optimize for increasing clicks and avoid showing non-relevant content to the user (click-bait). This enabled Digitec Galaxus to exploit popular trends while also exploring potentially better-performing trends. 

How Google Cloud helps

The newsletter context-driven personalization system was built on Google Cloud architecture using the ML recommendation training and prediction solutions available within our ecosystem. 

Below is a diagram of the high-level architecture used:

The architecture covers three phases of generating context-driven ML predictions, including: 

ML Development: Designing and building the ML models and pipeline 
Vertex Notebooks are used as data science environments for experimentation and prototyping. Notebooks are also used to implement model training, scoring components, and pipelines. The source code is version controlled in Github. A continuous integration (CI) pipeline is set up to automatically run unit tests, build pipeline components, and store the container images to Cloud Container Registry. 

ML Training: Large-scale training and storing of ML models 
The training pipeline is executed on Vertex Pipelines. In essence, the pipeline trains the model using new training data extracted from BigQuery and produces a trained, validated contextual bandit model stored in the model registry. In our system, the model registry is a curated Cloud Storage. 

The training pipeline uses Dataflow for large scale data extraction, validation, processing, and model evaluation, and Vertex Training for large-scale distributed training of the model. AI Platform Pipelines also stores artifacts, the output of training models, produced by the various pipeline steps to Cloud Storage. Information about these artifacts are then stored in an ML metadata database in Cloud SQL. To learn more about how to build a Continuous Training Pipeline, read the documentation guide.

ML Serving: Deploying new algorithms and experiments in production 
The training pipeline uses batch prediction to generate many predictions at once using AI Platform Pipelines, allowing Digitec Galaxus to score large data sets. Once the predictions are produced, they are stored in Cloud Datastore for consumption. The pipeline uses the most recent contextual bandit model in the model registry to evaluate the inference dataset in BigQuery and give a ranked list of the best newsletters for each user, and persist it in Datastore. A Cloud Function is provided as a REST/HTTP endpoint to retrieve the precomputed predictions from Datastore.

All components of the code and architecture are modular and easy to use, which means they can be adapted and tweaked to several other use cases within the company as well.

Better newsletter predictions for millions

The newsletter prediction system was first deployed in production in February, and Digitec Galaxus has been using it to personalize over 2 million newsletters a week for subscribers. The results have been impressive, 50% higher than our baseline. However, the collaboration is still ongoing to improve the results even more. 

“Working at this level in direct exchange with Google’s machine learning experts is a unique opportunity for us. The use of contextual bandits in the targeting of our recommendations enables us to pursue completely new approaches in personalization by also personalizing the delivery of the respective recommender to the user. We have already achieved good results in our newsletter in initial experiments and are now working on extending the approach to the entire newsletter by including more contextual data about the bandits arms. Furthermore, as a next step, we intend to apply the system to our online store as well, in order to provide our users with an even more personalized experience. To build this scalable solution, we are using Google’s open source tools such as TFX and TF Agents, as well as Google Cloud Services such as Compute Engine, Cloud Machine Learning Engine, Kubernetes Engine and Cloud Dataflow.”—Christian Sager, Product Owner, Personalization ( Digitec Galaxus)

Since the existing architecture and system is also dynamic, it will automatically adapt to new behaviours, trends, and users. As a result, Digitec Galaxus plans to re-use the same components and extend the existing system to help them improve the personalization of their homepage and other current use cases they have within the company. Beyond clicks and user engagement, the system’s flexibility also allows for future optimization of other criteria. It’s a very exciting time and we can’t wait to see what they build next!

In the past decade, the Healthcare and Life Sciences industry has enjoyed a boon in technological and scientific advancement. New insights and possibilities are revealed almost daily. At Google Cloud, driving innovation in cloud computing is in our DNA. Our team is dedicated to sharing ways Google Cloud can be used to accelerate scientific discovery. For example, the recent announcement of AlphaFold2 showcases a scientific breakthrough, powered by Google Cloud, that will promote a quantum leap in the field of proteomics. In this blog, we’ll review another omics use case, single-cell analysis, and how Google Cloud’s Dataproc and NVIDIA GPUs can help accelerate that analysis.

The Need for Performance in Scientific Analysis

The ability to understand the causal relationship between genotypes and phenotypes is one of the long-standing challenges in biology and medicine. Understanding and drawing insights from the complexity of biological systems abounds from the actual code of life (DNA) through to expression of genes (RNA) to translation of gene transcripts into proteins that function in different pathways, cells, and tissues within an organism. Even the smallest of changes in our DNA can have large impacts on protein expression, structure, and function, which ultimately drives development and response – at both cellular and organism levels. And, as the omics space becomes increasingly data- and compute-intensive, research requires an adequate informatics infrastructure. An infrastructure that scales with growing data demands, enables a diverse range of resource-intensive computational activities, and is affordable and efficient – reducing data bottlenecks and enabling researchers to maximize insight. 

But where do all these data and compute challenges come from and what makes scientific study so arduous? The layers of biological complexity begin to be made apparent immediately when looking at not just the genes themselves, but their expression. Although all the cells in our body share nearly identical genotypes, our many diverse cell types (e.g. hepatocytes versus melanocytes) express a unique subset of genes necessary for specific functions, making transcriptomics a more powerful method of analysis by allowing researchers to map gene expression to observable traits. Studies have shown that gene expression is heterogeneous, even in similar cell types. Yet, conventional sequencing methods require DNA or RNA extracted from a cell population. The development of single-cell sequencing was pivotal to the omics field. Single-cell RNA sequencing has been critical in allowing scientists to study transcriptomes across large numbers of individual cells. 

Despite its potential, and the increasing availability of single-cell sequencing technology, there are several obstacles: an ever increasing volume of high-dimensionality data, the need to integrate data across different types of measurements (e.g. genetic variants, transcript and protein expression, epigenetics) and across samples or conditions, as well as varying levels of resolution and the granularity needed to map specific cell types or states. These challenges present themselves in a number of ways including background noise, signal dropouts requiring imputation, and limited bioinformatics pipelines that lack statistical flexibility. These and other challenges result in analysis workflows that are very slow, prohibiting the iterative, visual, and interactive analysis required to detect differential gene activity. 

Accelerating Performance

Cloud computing can help not only with data challenges, but with some of the biggest obstacles: scalability, performance, and automation of analysis. To address several of the data and infrastructure challenges facing single-cell analysis, NVIDIA developed end-to-end accelerated single-cell RNA sequencing workflows that can be paired with Google Cloud Dataproc, a fully-managed service for running open source frameworks like Spark, Hadoop, and RAPIDS. The Jupyter notebooks that power these workflows include examples using samples like human lung cells and mouse brains cells and demonstrate acceleration between CPU-based processing compared to GPU-based workflows. 

Google Cloud Dataproc powers the NVIDIA GPU-based approach and demonstrates data processing capabilities and acceleration, which in turn have the potential of delivering considerable performance gains.  When paired with RAPIDS, practitioners can accelerate data science pipelines on NVIDIA GPUs, reducing operations like data loading, processing, and training from hours to seconds. RAPIDS abstracts the complexities of accelerated data science by building upon popular Python and Java libraries effortlessly. When applying RAPIDS and NVIDIA accelerated compute to single-cell genomics use cases, practitioners can churn through analysis of a million cells in only a few minutes.

Give it a Try

The journey to realizing the full potential of omics is long; but through collaboration with industry experts, customers, and partners like NVIDIA, Google Cloud is here to help shine a light on the road ahead. To learn more about the notebook provided for single-cell genomic analysis, please take a look at NVIDIA’s walkthrough. To give this pattern a try on Dataproc, please visit our technical reference guide.

Over the past few decades, organizations have mastered the science of data warehousing. They have increasingly applied descriptive analytics to large quantities of stored data, gaining insight into their core business operations. Conventional Business Intelligence (BI), which focuses on querying, reporting, and Online Analytical Processing, might have been a differentiating factor in the past, either making or breaking a company, but it’s no longer sufficient.

Today, not only do organizations need to understand past events using descriptive analytics, they need predictive analytics, which often uses machine learning (ML) to extract data patterns and make probabilistic claims about the future. The ultimate goal is to develop prescriptive analytics that combine lessons from the past with predictions about the future to automatically guide real-time actions.

Traditional data warehouse practices capture raw data from various sources, which are often Online Transactional Processing (OLTP) systems. Then, a subset of data is extracted in batches, transformed based on a defined schema, and loaded into the data warehouse. Because traditional data warehouses capture a subset of data in batches and store data based on rigid schemas, they are unsuitable for handling real-time analysis or responding to spontaneous queries. Google designed BigQuery in part in response to these inherent limitations.

Innovative ideas are often slowed by the size and complexity of the IT organization that implements and maintains these traditional data warehouses. It can take years and substantial investment to build a scalable, highly available, and secure data warehouse architecture. BigQuery offers sophisticated software as a service (SaaS) technology that can be used for serverless data warehouse operations. This lets you focus on advancing your core business while delegating infrastructure maintenance and platform development to Google Cloud.

BigQuery offers access to structured data storage, processing, and analytics that’s scalable, flexible, and cost effective. These characteristics are essential when your data volumes are growing exponentially—to make storage and processing resources available as needed, as well as to get value from that data. Furthermore, for organizations that are just starting with big data analytics and machine learning, and that want to avoid the potential complexities of on-premises big data systems, BigQuery offers a pay-as-you-go way to experiment with managed services.

With BigQuery, you can find answers to previously intractable problems, apply machine learning to discover emerging data patterns, and test new hypotheses. As a result, you have timely insight into how your business is performing, which enables you to modify processes for better results. In addition, the end user’s experience is often enriched with relevant insights gleaned from big data analysis, as we explain later in this series.

The migration framework

Undertaking a migration can be a complex and lengthy endeavor. Therefore, we recommend adhering to a framework to organize and structure the migration work in phases:

1. Prepare and discover: Prepare for your migration with workload and use case discovery.
2. Assess and plan: Assess and prioritize use cases, define measures of success, and plan your migration.
3. Execute: Iterate the following steps for each use case:
   1. Migrate (offload): Migrate only your data, schema, and downstream business applications.
   2. Migrate (full): Alternatively, migrate the use case fully end-to-end. The same as Migrate (offload), with the addition of the upstream data pipelines.
   3. Verify and validate: Test and validate the migration to assess return on investment.

The following diagram illustrates the recommended framework and shows how the different phases are connected:

For a deeper understanding, read Migrating data warehouses to BigQuery: Introduction and overview