GRIFFEYE AT WORK
“Much of the work in CSA investigations is about finding the needle in the haystack”
Creating advanced solutions that create a meaningful impact for our customers is not a one-and-done achievement, but requires constant contact and collaboration between our research and development (R&D) teams and the law enforcement community.
John Lundberg is one of the Software Engineers and Team Lead on the Brain Team at Griffeye, which is dedicated to working on AI and Computer Vision. He completed his Master’s thesis at Griffeye ten years ago, and has been working in various roles here ever since. According to John, customer feedback is of critical importance to carrying out the team’s mission. Here, we chat with him to gain a closer perspective on how the collaboration and feedback process between Griffeye’s R&D and law enforcement works.
Why is it so important for Griffeye to work closely with our customers?
Our customers are law enforcement officers who investigate child sexual abuse (CSA) cases, and the material that investigators face is illegal to possess, and extremely harrowing to watch. We are all deeply motivated to help them in their work. But since the kind of work they do is not something we developers have first-hand experience in, we need to talk to them in order to truly understand the unique challenges they face.
We ask our customers to test the solutions we develop in their own environments using real world data we don’t have access to. Due to the distressing nature of their work, we also ask for their feedback on ways we can help improve their mental wellbeing through different kinds of features.
What does the communication between the Brain team and customers look like?
We collect feedback in a number of ways. For example, there are several current and former law enforcement officers who teach our software in training courses. We work closely with them to get feedback on new features. They also help us capture feature requests and feedback from the users they meet as part of the training.
For larger projects, we set up user group sessions where we present our ideas and share early versions to get feedback at an early stage. We also gain a lot of insights from our user forum, where users post both general questions and feature requests.
Can you give us an example of a challenge that your team has tackled, and how you solved it?
Much of the work in CSA investigations is about finding the needle in the haystack. And this is one of the primary problem areas our team works with every day. Seized material can contain millions of images and videos, and we have a few different solutions to help investigators quickly find any CSA material that might be there.
For example, of particular concern is any new material that has not been previously seen in a CSA investigation. New material can indicate that there is a child that is currently being abused by a perpetrator or by someone that the perpetrator has been in contact with.
To locate the CSA material that has not been seen before, the Brain Team has worked in close collaboration with several national and federal law enforcement agencies to develop a CSA image classifier. Using machine learning, we’ve trained the CSA image classifier with known CSA images to help it discover new images that likely also depict CSA.
This classifier can be used to prioritize which files should be looked at first. Videos are especially distressing for investigators to review, so we also made it possible to use the image classifier to scan videos and indicate which frames likely depict CSA. That way, investigators can quickly look at those frames and reduce the time needed for review.
How did that create impact for our customers?
The CSA image classifier helps investigators quickly track down the material that is most relevant for victim identification. At the same time, since the whole review process is shortened, it can also significantly reduce exposure time to distressing material
Thank you for sharing your experience with us, John!
GRIFFEYE AT WORK is a series of interviews with employees exploring topics related to life at work, how we develop our products—and who the people are behind them.