Dying for that bag Kerry Washington was toting on last week’s Scandal? Need the sweatshirt you saw Kanye wearing on Instagram? LookLive has you covered. The Amsterdam-based startup spent two years developing tech that identifies exact items of clothing worn by your favorite characters and stars. CEO Christopher Archer explains how they did it.

Companies have been promising shoppable TV for years. What’s been the holdup?

The technical integration is easy peasy, but the challenge is the metadata, and how you get it. If I’m watching The Voice, how do I know what jacket Pharrell is wearing? The broadcaster and network don’t know -- they’ll tell you to call the costume designer. It’s just impossible to do this at scale. When we started, we said that if we can’t figure out how to identify items independently, we’re going to shut down.

So what was the magic trick?

We use a mix of image-recognition technology and machine learning. People have what we call a style profile -- essentially the identity of self. Claire Underwood is going to dress in a specific way from specific brands in a specific price range. The machine-learning tech learns this, and when it sees a white three-button blazer on her, it knows she would never wear H&M or Forever 21, but she would wear Prada and Armani. That’s where the narrow search for product identification begins.

Users can purchase clothing from your mobile app and your website. Are we inching toward real-time shopping right on the TV screen?

The challenge is still speed and scalability. Right now, we can have every TV show produced and ready to shop within three hours of its airing. But we have our fingers crossed that we’ll have it fully automated in real time within the next 12 to 18 months.

Check out more companies on the 2016 Brilliant 100 list.