AI with a purpose: Behind the scenes at Media Distillery

When Adrián Arroyo joined Media Distillery in January 2019 his role was to coordinate the team of professionals who trained the company’s automated video analysis algorithms to accurately detect the start and end times of programmes in a TV live stream. The launch of ChatGPT was still almost three years away and Artificial Intelligence wasn’t quite the buzzword it is today. Most members of the public still thought of AI as something from the movies. Today, both the company and Adrián’s role have changed pretty dramatically, developing and embracing the use of AI for practical purposes.

“Although everyone is now talking about AI, from a lot of companies it still feels like ‘smoke and mirrors’ - using AI so they can say they’re using AI," says Adrián. “That’s definitely not the case at Media Distillery. We started out using Machine Learning and then over time focused on specific forms of AI, like generative AI, because they are the best way to solve the problems our customers face. It’s about adding value: AI with a purpose."

Collaboration with customers is key

As a product owner for Time Marker Suite, which includes EPG Correction Distillery, and Ad Break Distillery, Adrián works closely with Media Distillery’s customers to further that goal of adding value. “Previously, my main contact with our Pay TV and OTT clients was during their onboarding process, making sure they were up and running on our solutions,” Adrián explains. “But now I’m meeting with them on a regular basis to understand their needs and consider new functionality and use cases that we could support with our solutions and future development. It’s great to have that direct contact to help shape our development.”

Speed is of the essence for replay user experience

So what’s the top priority on his product wish list? “We’re always looking for ways to improve our solutions in line with our strategic goals,” says Adrián. “Speed is a big factor. We want our customers to get the results of our video analysis in as close to real time as possible so they can start using it as soon as possible. With products like EPG Correction Distillery and Ad-Break Distillery, we’re aiming to improve the replay user experience and we know that the vast majority of replay viewing happens within minutes or hours of the original broadcast, so we are continuously tweaking the solution to optimize for speed.”

A constant focus on accuracy in metadata generation

Adrián also stresses that accuracy is a constant focus. “The beauty of AI models is that they can become smarter all the time - if you train them correctly. So, we’re working on a range of new training tools that will help us to continue improving the accuracy of the metadata we generate for our customers, where we more precisely identify events in a video stream like the start and end of a program or ad break.” This is particularly important for the next phase of the Ad Break Distillery solution: “You need pinpoint accuracy if you’re going to replace the adverts that ran in the linear program with different ads for the replay version. In this scenario, you can’t run the risk of showing even a few seconds of one advert before another starts - that’s not acceptable to the advertiser and revenue would suffer. It’s something we’re making real progress with at the moment with our in-house ML and AI experts.”

The shift to cloud-based solutions

Behind the scenes, there’s also plenty of work going on to reduce overheads - both time and money - for Media Distillery and its customers. “Since early 2023 we’ve been working hard to switch from an on-premises deployment model to a cloud-based solution,” Adrián confirms. “We now have a number of newer customers who’ve been onboarded directly into our cloud-hosted solution which is a much faster and more scalable deployment option because the customers don't have to host additional hardware in their data center and provide support for it. Now it’s possible for customers to start getting AI-generated metadata from their video streams within just a few days.”

For those customers who’ve been with Media Distillery for many years, there’s a process of migration to the cloud. “We’ve worked hard to develop our own tooling to manage the migration for those who want to enjoy the greater flexibility, resilience and scale offered in the cloud. There really weren’t any tools there to support our needs for ingestion of live streams, but now we’ve developed them the migration process is really smooth and simple.”

Media Distillery is a place to grow and develop

Although originally from Spain, Adrián has lived and worked in Amsterdam for a decade and enjoys the flexibility to combine working from home and from Media Distillery’s historic office building in the city center. “From a personal standpoint, one of the best things about Media Distillery is the opportunity to grow my skills and try out different roles,” said Adrián. “From where I started as a coordinator for the support team, to where I am today as Senior Product Owner, is quite a big change in just five years. As the company has grown, they’ve given me the chance and space to develop and learn.”

Adrián had previously worked for larger organizations, including metadata experts Gracenote. “Moving to a smaller start-up could have been seen as a risky move,” he admits. “But the team dynamics at Media Distillery are great and it’s a very collaborative environment, both between ourselves and with our customers. It’s also a very flat hierarchy. The management are willing to try new things and take on board ideas from the team, which is great for agility but also for motivation and job satisfaction.”

Want to know more?

If you’d like to learn more about how Media Distillery’s solutions could help your Pay TV, broadcast or OTT business to delight viewers with innovative user experiences, contact us for a demonstration.

Or learn more about career opportunities with Media Distillery.