Owlin: Helping engineers get more sleep with automated infrastructure on demand

About Owlin

Based in the Netherlands, Owlin delivers big data and AI solutions for clients including ING Bank, Adyen, Deloitte, and KPMG. Its lead product analyzes millions of news sources to give finance professionals uniquely customized, near real-time data on global risks and trends.

Industries: Financial Services & Insurance
Location: Netherlands

About Incentro

Incentro is a full-service digital agency and consultancy with offices in the Netherlands, Spain, and Kenya.

When Owlin, the financial news analysis platform, saw that managing servers was keeping engineers up at night, it switched to an automated Google Cloud alternative to help them get more sleep.

Google Cloud results

  • Deploys to new client regions in one week, instead of six to 12 months
  • Reduces costs, thanks to automated deployments of preemptible VMs
  • Generates infrastructure for new translation services in seconds, instead of weeks

Scales automatically to meet 20X increase in user base

Navigating the ups and downs of financial markets is more than just a numbers game. Following global news events, from local politics to the weather, is vital for any financial professional looking to make informed investments. Owlin helps them do exactly that, with a news analytics platform accessible through a dashboard, API, or integrated widgets. By drawing on more than 2.8 million sources in 11 languages, Owlin transforms the news into usable intelligence in near real time for clients including Fitch Ratings, ING Bank, KPMG, Philips, Deutsche Bank, and Adyen.

“We scrape news from big news sites to smaller blogs from all around the world,” explains Sjoerd Leemhuis, co-founder and CEO of the company. “Then we translate all of those articles into English and query them with algorithms that detect risks and opportunities for our customers. We don’t just give them the news: we visualize metadata to give them an authoritative overview on trends and developments, aggregated from multiple sources. We can compare the number of bankruptcies taking place in a certain country’s telecom industry and put it in the context of historical tendencies, for example. Or we can check if your portfolio indirectly harms the environment, or is linked to child labor or unequal pay.”

“One of our KPIs is how much sleep we each get. Engineers responsible for infrastructure used to report less than six hours sleep a night, due to the stresses of the job. Since we moved to Google Cloud, that’s no longer the case. Sleep, happiness, and productivity are all up.”

Sjoerd Leemhuis, co-founder and CEO, Owlin

Owlin originally used on-premises servers for infrastructure, but as its client base expanded, limitations surfaced. “Uptime is extremely important to our customers, so we have to keep our infrastructure up 24/7,” says Sjoerd. “A resilient architecture needs to have failover capacity, but doubling up infrastructure means more than just buying extra servers. There are logistical issues around installation and maintenance, for example. We had no way to create new servers overnight, but with demanding clients in a fast-moving industry, that’s something you have to be ready to do.”

Beyond commercial and technical issues, Sjoerd and the team were even more concerned about the human impact of maintaining infrastructure.

“One of our KPIs is how much sleep we each get,” says Sjoerd. “Engineers responsible for infrastructure used to report less than six hours sleep a night, due to the stresses of the job. Since we moved to Google Cloud, that’s no longer the case. Sleep, happiness, and productivity are all up.”

Managing preemptible VMs automatically for infrastructure on demand

Reliability is important, if not crucial, for any company delivering a digital service. As Owlin grew, it recognized stability issues inherent to its on-premises infrastructure. “We were running an ad hoc collection of servers co-located in a data center in Amsterdam,” says Sjoerd. “At first it was cost efficient, but we had problems, such as a local power outage which shut down our platform when the emergency generators failed. Most of all, maintaining it meant our engineers had to be ready round-the-clock to fix problems, which we didn’t feel was fair or sustainable.”

Owlin decided to migrate to the cloud and began evaluating major cloud providers, looking for a reputation that high-profile clients could trust. “Our evaluations showed that Google Cloud offered the best user experience, workflow, and services, as well as more competitive pricing for the resources provided,” says Sjoerd. “Google thinks far ahead about how to build and mature this kind of infrastructure, which makes sense. After all, it uses the same tools.”

With help from the consultancy Incentro, Owlin migrated its architecture to Compute Engine instances, including Preemptible VMs. “It turns out that the world doesn’t produce news at a constant rate,” says Sjoerd. “We need more servers during the week than we do at the weekend, for example, so we wrote a management layer that looks at the flow of news through our ingestion pipeline and dynamically spins up or shuts down preemptible VMs. Now, when we onboard new customers, we can quickly increase the amount of processing power available to our infrastructure. We use what we need, when we need it.”

“We don’t have to calculate how to provision hardware for peak capacity. With our software, we scale preemptible VMs automatically, cost-effective resources are added and removed as required, and it reassigns tasks when preemptible VMs expire, minimizing disruption. It’s all automated, so nobody needs to oversee it.”

Ralf Nieuwenhuijsen, Head of Software Engineering, Owlin

Owlin already used Elasticsearch to store and search datasets, but now integrates this with Cloud Bigtable as a more authoritative data source. “We operate a really big Elasticsearch cluster, but it’s not very reliable and can lose data during network partitions,” says Ralf Nieuwenhuijsen, Head of Software Engineering at Owlin. “We wanted a NoSQL database that we could trust as an authoritative, stable data store. We don’t always need the full range of features offered by Cloud Spanner, so Cloud Bigtable is a cost-effective alternative.”

For the team’s engineers, the combination of automated scaling of server capacity and reliable managed services means they can spend more time on product development, optimization, and of course, sleep.

“We don’t have to calculate how to provision hardware for peak capacity,” says Ralf. “With our software, we scale preemptible VMs automatically, cost-effective resources are added and removed as required, and it reassigns tasks when preemptible VMs expire, minimizing disruption. It’s all automated, so nobody needs to oversee it.”

Deploying new translation capability at speed with Cloud GPUs

Translation is a vital element of the Owlin platform, and the team uses Cloud GPUs on Compute Engine to dramatically increase the volume of material it sends through its translation algorithms. Owlin’s own algorithms are optimized for search, rather than human readability, so the platform uses Cloud Translation for customer-facing translation.

“Deploying a new language to our platform used to take weeks,” says Ralf. “We would need to buy, install, and set up five to 10 new servers in our data center, and then maintain them. With Compute Engine, we can provision everything we need for a new language in a couple of clicks. Now we’re adding a new language every couple of months.”

Reduced time to market for rapid global growth

In the past year, Owlin’s customer base has increased by a factor of twenty as it deploys its services to clients across the world. “Before we moved to Google Cloud, preparing for a new client in another part of the world would be a 6- to 12-month project,” says Sjoerd. “Now it only takes a week. Most of our clients are in the US, but thanks to Google Cloud, they don’t need to connect to a data center abroad. That makes a real difference in terms of latency for investment managers, who want immediate results when they click on something.”

“Time to market is really important in our business. Google Cloud has really reduced that and helped us focus on ways to improve our service, rather than maintain it. But for the team, the most important thing is peace of mind, knowing we have a solid architecture behind us.”

Sjoerd Leemhuis, co-founder and CEO, Owlin

Since migrating to Google Cloud, Owlin has been ready to scale up for new users and new demands at speed, without additional stress. When Fitch Ratings, the world’s third-largest rating agency, wanted to integrate Owlin into a web-based dashboard for an additional 16,000 users worldwide, Owlin was ready for the challenge. “In the past, that would have been something to panic about,” says Sjoerd. “Now we can take it in our stride.”

“Time to market is really important in our business,” adds Sjoerd. “Google Cloud has really reduced that and helped us focus on ways to improve our service, rather than maintain it. But for the team, the most important thing is peace of mind, knowing we have a solid architecture behind us.”

About Owlin

Based in the Netherlands, Owlin delivers big data and AI solutions for clients including ING Bank, Adyen, Deloitte, and KPMG. Its lead product analyzes millions of news sources to give finance professionals uniquely customized, near real-time data on global risks and trends.

Industries: Financial Services & Insurance
Location: Netherlands

About Incentro

Incentro is a full-service digital agency and consultancy with offices in the Netherlands, Spain, and Kenya.