Solving Logically' s Core Challenges using Google Cloud

Logically is a misinformation-fighting news and fact checker app that combines insights from AI and human intelligence to determine the credibility of digital content. It does so by recognizing biased language, charting the tone of coverage on a person or event, and then providing users with as much context as possible. Users can send content from social media platforms to be analyzed, and a team of fact-checkers at Logically sets the process in motion.

To bring this fact-checking machine to life, Logically needs to ingest as much information as possible from social media feeds and news channels to train its machine learning models. To achieve the high scalability this task demands in an economically viable way, Logically built its cloud-native infrastructure on Google Cloud, effectively using its managed services and a microservices framework since it was first launched, in April 2019.

Increasing the Throughput of Data and the Number of Concurrent Calculations Within the Application

Today, Logically has 30,000 daily active users. More than 200,000 articles and 10 million pieces of social media content are ingested and processed by Logically on a daily basis. To analyse each and every one of those articles, Logically uses Google Cloud GPUs (graphic processing units) hosted on Google Compute Engine to increase throughput of data and the number of concurrent calculations within the application.

To provide users with timely news on a broad range of topics, Logically also connects GPUs to Google Kubernetes Engine (GKE) to harness its managed services and auto-scaling capabilities. Meanwhile, using Google Cloud Storage, Google Cloud SQL, and TensorFlow Enterprise, the company ingests content from all over the world to keep the data on its platform up-to-date so that it can respond quickly to users' searches. More recently, Logically also set up Google Cloud TPU on GKE to accelerate its machine learning workloads.

Achieving Scalability using Google Preemptible Virtual Machines (PVMs) on Google Cloud

At the outset of COVID-19, there was an exponential surge in online content about the virus, including conspiracy theories. Logically' s workload suddenly multiplied tenfold as it supported public sector partners in fighting misinformation. In March 2020, to support Logically in dealing with the surge in its workload quickly and accurately, Google Cloud connected Searce, Google Cloud' s Global Premier Partner and a modern technology consulting firm that empowers clients to futurify their businesses leveraging Cloud, AI & Analytics.

To help Logically significantly increase its capacity by getting the right tradeoff between cost and throughput, Searce suggested the use of Google Preemptible Virtual Machines (PVMs) on Google Cloud: an alternative to regular compute instances that has a short life span and can be around 80% cheaper.

Because PVMs only run for 24 hours, Searce also helped Logically to implement best practices and patterns to leverage cost benefits while keeping the service reliable. With assistance from Searce, Logically was able to leverage PVMs on GKE. Building on experience, Searce enforced countermeasures for ensuring reliability and availability of PVM nodes to offset the fickle nature of PVMs while reaping the benefits from the discounted rates. The project was running successfully within two weeks.

Additionally, Searce performed a security audit of Logically' s infrastructure and networking on Google Cloud and implemented a checklist of security policies and measures to be enforced and maintained for tightening and securing Logically' s infrastructure on Google Cloud. These recommendations included changes to the organization structure and project/environment hierarchy.

The Growing Business Impact and Google Cloud Solutions

With Google Preemptible Virtual Machines, Logically scaled its infrastructure to a capacity 10 times higher than its usual, so that partners could keep relying on the app to inform their policies in a time of crisis.

The rapid scaling and lack of availability issues contributed to Logically finding tens of thousands of instances of misinformation relating to COVID-19, which may otherwise have gone undetected. It also helped the company to recommend countermeasures to public-sector partners that rely on the app to verify content which informs their policies, helping them to protect their hundreds of millions of followers from misinformation.