Google Cloud Results
- Trains custom-designed machine learning algorithms on ever-expanding datasets with Compute Engine
- Saves time and money with managed services, freeing engineers to focus on the product
- Collaborates with internal teams easily and more securely using BigQuery to share data
The introduction of electronic medical records (EMRs) over the last ten years has transformed critical care units in hospitals all over the world. A decade ago, doctors and nurses would take between ten and twenty measurements for each patient and charted out their conditions by hand. Today, EMRs enable hospitals to use sensors and machines to take up to 100,000 measurements per patient per minute, providing much more accurate and detailed information.
However, as hospitals collected more and more data, new challenges emerged. Medical professionals would find themselves overwhelmed with data, giving rise to a new range of clinical analysis tools designed to help separate the signal from the noise. For CLEW, this was the perfect opportunity to apply its machine learning expertise.
“Our initial cloud infrastructure was built in a way that was expensive and did not give us the flexibility we needed. With Google Cloud Platform, we found the right combination of scalability, availability, and affordability.”
—Izik Itzhakov, VP Business Development, CLEW
“Generating data is not the problem,” says Izik Itzhakov, VP Business Development at CLEW. “When we spoke with people in the industry, they said that they wanted to be able to look at the right data and make sense of it.”
Since 2015, Israel-based CLEW has been providing hospitals with an analytics platform that uses cutting edge machine learning algorithms to help prioritize treatment for the patients who need it most. In late 2016, as the company expanded, its infrastructure began to show signs of stress with stability issues and valuable time spent on maintenance. CLEW needed a new infrastructure that could handle high-speed, high-volume analytics, and keep resources costs at a minimum. It turned to Google Cloud Platform (GCP).
“Our initial cloud infrastructure was built in a way that was expensive and did not give us the flexibility we needed,” says Izik. “With Google Cloud Platform, we found the right combination of scalability, availability, and affordability.”
Stability in the cloud with Google Compute Engine
CLEW sits at the intersection of two revolutionary movements in two separate industries. The first was the growth of EMRs in medicine. The second was the introduction of machine learning in the world of data science.
Seeing an opportunity to use machine learning to solve the new challenges of EMRs, CLEW partnered with pioneering medical institutions across Israel and the United States, including the prestigious Mayo Clinic in Rochester, Minnesota. CLEW’s goal was to help doctors diagnose patients in critical care faster and more accurately than ever before. To do that, the company built a warehouse of anonymized patient data. With 10 billion data points from approximately 200,000 anonymous patients, the company trained its algorithms to spot patterns and feed insight back to its platform, providing real-time, data-driven insight to busy doctors.
In its early days, CLEW started out with a single on-premises server. As its dataset grew, operations took longer, sometimes days, to complete, and the company moved to a leading cloud provider. “We needed to move at a much faster pace and did not want to keep spending money to continuously build on the internal infrastructure,” says Izik.
“Our engineers found Compute Engine very easy to use, ramping up machines when we needed to, turning them off when we were done with them. It made the overall experience much smoother.”
—Avigdor Faians, VP Product and Co-Founder, CLEW
The new cloud infrastructure worked well at first, but as the company continued to scale it struggled to keep up without spending vital resources on it. “One of our key engineers would be called off projects to configure servers and improve stability,” says Avigdor Faians, VP Product and Co-Founder at CLEW. “It was a waste of his time.” With infrastructure costs mounting and its developers becoming overstretched, CLEW looked for a new solution.
The company partnered with MORE to build its new infrastructure with Google Cloud Platform (GCP). “MORE provided us with expertise on the architecture and how best to optimize it,” says Avigdor. “We don’t have a dedicated DevOps staff, so MORE’s cooperation was key for the transition.” With MORE’s help, CLEW migrated to GCP with minimal fuss in less than a month.
The new infrastructure was built with Compute Engine, which gave CLEW the convenience of an intuitive interface and clear pricing along with the flexibility to use the company’s own proprietary machine learning tools. “We do a lot of custom design for our machine learning so we don’t use any off-the-shelf products for our algorithms,” says Avigdor. “We need the freedom to be agnostic to whichever cloud environment we are using.”
Meanwhile, the company used Cloud Storage to hold its masses of training data in an accessible, scalable way. Sometime after the migration, CLEW began using BigQuery to provide anonymized slices of data to doctors or technical staff at hospitals who could easily run queries without compromising the security of the whole dataset.
“Our engineers found Compute Engine very easy to use, ramping up machines when we needed to, turning them off when we were done with them,” says Avigdor. “It made the overall experience much smoother.”
“The overall cost is not just the monthly bills. If an engineer wastes time on configuration, that’s quite expensive for me. Now they can be working on the product or the algorithms.”
—Avigdor Faians, VP Product and Co-Founder, CLEW
Greater capacity and better allocated resources
With Google Cloud Platform, CLEW built an infrastructure that not only scales with the company’s success but also provides significant time and money savings. “Our infrastructure bills are around one third of what they were with our previous platform,” says Izik. But Avigdor points out that the savings only start there. “The overall cost is not just the monthly bills,” he says. “If an engineer wastes time on configuration, that’s quite expensive for me. Now they can be working on the product or the algorithms.”
With the migration to GCP complete, CLEW is looking to expand the capabilities of its analytics platform beyond the critical care units. The company is exploring options to add a new layer of security protocols to comply with data security regulations, such as HIPAA, allowing it to work with unstructured text such as doctors’ notes and improving its predictive powers even further. Meanwhile, CLEW sees an opportunity with Google Cloud to have more than just its machine learning capabilities on the cloud.
“We currently have our product installed on-premises in hospitals but the next big thing for us with Google, is to have a more secure cloud,” says Izik. “That will help us scale faster, deploy more quickly, and merge datasets. It’s important for us to get the security right from a technical perspective and to reassure the hospitals that their data is safe in the cloud.”