Data and Machine Learning

Using AI and AWS Cloud, Akrivia Helps Researchers Fight Alzheimer’s

Trek10 and Akrivia work together to prepare Akrivia's secure database for use by mental health researchers

Akrivia Health began as a small research project in Oxford University, but over the past ten years it has grown into an independent company with a hard-hitting mission: change the trajectory of mental health research as we know it. Through Akrivia’s data platform, mental health researchers will have access to datasets enriched to a degree that has never before been available for study, paving the way for swifter diagnoses and better treatment options for patients with conditions such as Alzheimer’s, dementia, and depression.

In the UK health system, patient symptoms and other invaluable information is often stored as unstructured data, written into note fields in charts. As such, it has previously been all but inaccessible at scale. Akrivia developed natural language processing AI that can parse that unstructured data and pull out important insights, partnering with the NHS, Britain’s National Health Service, to turn that raw data into an anonymized, research-ready dataset. Currently, Akrivia’s dataset has over 3 million records and is more comprehensive than anything ever made available to researchers in the UK.

“We wanted to create an application that was more functional, fluid, and usable for research and industry applications than anything that’s ever been done,” said Benjamin Fell, Head of Research at Akrivia.

But to use the data, researchers would need an easy and secure way to access it. Akrivia set out to create a research-friendly web app, which meant moving the data off their data center and putting it in the cloud—far from trivial, considering the sky-high security requirements involved in handling sensitive patient information with the NHS.

“Our biggest concern at all times is security and governance,” said Andrew Fry, Head of Data Engineering at Akrivia. “Getting expert review and advice from Trek10 ensured confidence in our new platform.”

Akrivia met with Trek10’s EU team in August, planning to launch their research application by April of the following year. Together, the two teams got to work designing an architecture that was air tight enough to pass NHS security standards, and a revolutionary data tiering system that would expand researcher access to life-changing health data.

Choosing the Cloud Provider

“Our working relationship all started out as a healthy debate over LinkedIn, actually,” said Nessan Harpur, Client Success Manager at Trek10.

With Akrivia on the cusp of a cloud transition, CEO Mike Denis had already begun to weigh the pros and cons of each provider: should they go with AWS, Azure, IBM? “Mike and I just chatted about the benefits and costs associated with each platform and ultimately,” Harpur said, “Akrivia decided that AWS was the best provider for the job.”

The frankness of those early conversations and the close presence of Trek10’s EU team earned Akrivia’s trust, and the two teams joined together to hone in on Akrivia’s AWS architecture and security practices.

But the opportunity for Akrivia went deeper than the cloud architecture alone. The way they planned to tier and anonymize the data could significantly widen its research possibilities.

Color-coding the Data

Typically, the NHS would have a long approval process for researchers who wanted access to data like this, due to the sensitivity of patient information it included. Akrivia’s goal with their platform was to anonymize the raw NHS data in a way that enabled a standardized, streamlined approval process. This would allow more researchers to conduct more studies, widening the possibilities for pharmaceuticals and treatments.

Alongside Trek10, Akrivia brainstormed how to handle the various tiers of data, some of which contained PII and PHI (Personally-Identifiable Information and Patient Health Information, respectively) that had to remain completely inaccessible to anyone along the chain.

“What we came up with is truly what sets our platform apart,” said Fry. “By solving the technical challenge of secure data governance at each stage, you get something that can be accessed far more readily.”

Designing an NHS-worthy AWS Infrastructure

Trek10 and Akrivia decided the infrastructure itself would run on RDS, Fargate, and EKS, using Elastic Beanstalk for initial deployments for the sake of speed. A multi-account structure would help with account governance and role-based access control would make Akrivia’s deployments safer. Trek10 also laid the groundwork for logging, monitoring, and alerts, getting set up with CloudWatch logs and Kinesis Data Firehose.

“We knew AWS had managed services for everything but we didn’t know which would be best for us,” said Andrew Fry. “Trek10 talked through every major decision with our team and were great at pushing us in the right direction.”

“Akrivia is an inspiring company to be involved with,” added Trek10’s Nessan Harpur. “They are directly impacting so many people’s lives.”

Race to the Finish Line

The backend was solidified by early December, but there was a lot to do to make the application production-ready. From Akrivia’s side, they were testing and bug fixing the application, getting ready for pen testing, all of their gears spinning at once.

“We were on a really tight timeline,” said Fry, “and Trek10 were there with us on Slack probably once a week when things popped up.”

With less than six weeks left to the scheduled launch, Fry reached out to Trek10 with a request: a Kubernetes deployment was halfway done, and they needed it put on the cloud on EKS using RDS, and their hands were increasingly full. Could Trek10 take it over? “I remember they just said, ‘Yep, we can do that.’” said Fry. “It’s been fantastic for us to have their help like this on whatever we throw at them so close to the finish line.”

Changing the Future of Mental Health

Akrivia’s main goal now and into the future is to increase the research done on their datasets, and help their NHS partners, as well as academics, access it effectively and regularly. They’re actively improving data upload frequency and changing updating their data ingestion so the data is always up to date and accessible.

“For a platform like ours, technical infrastructure goes hand in hand with information governance“, said Head of Research Benjamin Fell. “The architecture we’ve designed with Trek10 has been crucial for ensuring that researchers can use the data to improve mental health care in a way that is secure, legal and ethical.”

As Akrivia looks to their new platform on the cusp of launch and everything that lies ahead, they are excited for what the coming years hold for the future of mental health.

“We’ve got so much data now”, says Andrew Fry, “I can already think of half a dozen research projects that would be great to write up—studies that literally, as little as a few months ago, wouldn’t have been possible. Now they are.”