Apr 15, 20203 min read

How a Klarna engineer #hackedthecrisis.


by Klarna

Meet our engineer Deepa Krishnamurthy, who’s team won in the category Save Lives – Digital solution in the Swedish Hack the Crisis hackathon!

Last week, the Swedish Government arranged an official Covid-19 hackathon with the aim to solve one of the biggest crises of our lifetime. Several of Klarna’s employees participated to find creative ideas and digital solutions for saving lives, businesses and communities. Among them was Deepa, who did an excellent job using her previous experience with machine learning to detect Covid-19 symptoms.

Around 7,000 individuals participated in the event, working on different solutions to fight the COVID-19 pandemic, so there were a lot of teams to choose from. Deepa decided to join the VoiceMed team, which had originated their idea two weeks earlier in Hack the Crisis Finland.

“I liked how serious it was, so that we could really contribute something.”

Tackling the lack of testing

Over the weekend they worked together on a solution to use machine learning and speech processing to detect Covid-19 symptoms with people’s voices. The team had previously developed a cough classification model to classify wet and dry cough with around 70% accuracy and Deepa was tasked with improving the model. Being an engineer and former data scientist who previously worked with developing machine learning models to learn payment behaviours, this was something that she was well-qualified to do and had exactly the skills the team needed to get something up and running quickly.

“The data here is very challenging so it wasn’t a simple model.”

During the weekend Deepa successfully worked on the model and was able to get to an accuracy of 84% – a huge improvement. This was the final result and the prototype that the team pitched in the hackathon – and which led to them winning their track, Save Lives – Digital solution.

“The Klarna guiding principle ‘start small, learn fast’ really helped!”

Aiming for market launch

Deepa is not stopping here. The team is going to continue working in their free time and will focus on getting a feasible product live within the next months. Deepa has taken over the lead role for the machine learning team to help channelise the team, build a roadmap, talk to different stakeholders and hopefully lead the way to releasing a machine learning product. They are hoping to get more collaborations from other startups and labs which are working on similar solutions, including ongoing talks with Cambridge University who are also looking into cough samples for Covid diagnosis.

“It is going to be really exciting to take it from an idea to a product. I hope we can make a difference with what we are doing.”

Congratulations to Deepa and her team – very impressive!

Read more about their solution here.

The jury’s motivation:

This service takes a huge strain off the existing medical support services and tackles the lack of testing available, by providing an ingenious yet simple technical solution that can easily be used by anyone of any age, and regardless of access to the internet.