HackWeek in the Age of COVID-19: Making the Best Out of the New Normal
By Johanna Fitschen
Our HackWeek offers all Scouts, techie or not, the opportunity to spend a week each year to develop ideas, pitch them in front of colleagues and work together as #OneTeam with Scouts from all departments to follow up on projects they have always wanted to tackle. This year, HackWeek was held entirely online. In different categories our Scouts can take different approaches with their ideas during HackWeek.
In this year’s category “Bleeding edge technology”, they had the chance to become creative with the help of state-of-the-art technology. As one of the winning teams, they can donate 500 € to a charitable organisation of their choice. The team consisting of Can Rager (Working Student Insights & Analytics), Dierk Bachmann (Senior Data Engineer), Ismail Benbrahim (Senior Data Analyst), Marc Than Van Con (Junior Data Analyst), Siddharth Raina (Data Scientist), Sinisia Milojevic (Senior Software Engineer) and Sumanth Badadal (Software Engineer) has won with their idea “Help our customers choose the right profile picture”. On behalf of the whole team, we talked to Ismail to find out what this idea is all about.
Hello Ismail, why did you take a closer look at our customer’s profile pictures during this year’s HackWeek?
Ismail: We focused on the interests of our customers. The team’s goal was to quantify those characteristics that make a profile photo appealing and trustworthy and use them to predict the conversion rate of a photo, in order to help our customers choose the right profile photo. The technical implementation uses the proven method called transfer learning. As time is of the essence during the HackWeek, we focused our efforts on only one neural network pretrained model called ResNet which we adapted to our particular task.
What happens to this project after HackWeek?
Ismail: We have identified and quantified how some key characteristics almost always translate into a higher conversion rate. We will share those findings with our customers. For example, we found out that for a male agent, wearing a complete suit is better than just a shirt. In addition to these communications, we will continue collecting more data. The data will be incorporated into the Machine Learning Model to improve its accuracy. This would allow to give our customers an initial evaluation of the image quality while uploading their profile photo.
What are you going to do with the prize money?
Ismail: My team and I will donate 500 EUR to Deepwave e.V., an organisation that has been dedicated to protecting the oceans for more than 15 years.
This project sounds cool to you? Why not joining us to realise similar projects? Here are some job offers that might be interesting for you:
