TCNJ

37 Class Notes SUMMER 2026 Artificial intelligence for your ears Raphael Rezkalla '17 is training the algorithm that knows your next favorite song. BILL CARDONI IF YOU'VE EVER MARVELED at a Spotify recommendation — a song matching your taste so well that it's like the music app knows you — it's likely Raphael Rezkalla '17 had a hand in it. "The things we're doing at Spotify, nobody in the industry has done yet," says Rezkalla, a senior artificial intelligence product manager. The cutting-edge technology driving Spotify's AI makes sense of all the input that listeners provide while using the platform. He says that whereas Facebook and Instagram users join communities that signal their interests, Spotify relies on behavioral data. It analyzes how long people listen to certain artists, which songs they play all the way through, and which they skip, to infer what they might enjoy next. Rezkalla's work in training large language models — the data-gathering and processing systems at the heart of what we know as AI — uses listener input, including selected songs and created playlists, to predict future preferences. He has also contributed to Spotify's Taste Profile project (see next page), which allows users to influence the algorithm by weighing in on whether Spotify's personalization attempts hit or miss the mark. "We're taking this relatively new technology and trying to apply it differently. Nobody in the industry knows if it's going to work; We're going along for the ride." REZKALLA HONED HIS INTEREST in large language models as a computer engineering major in college. A TCNJ adjunct professor helped him secure his first internship at the innovative ORBCOMM, a company dedicated to ways to collect and share data. "SpaceX was one of their first customers," he says. Rezkalla's path to Spotify included an internship at Bloomberg, which he landed through a TCNJ job fair. He spent six years at Meta, working on both Facebook and Threads, and then did a stint at Palantir right before he joined Spotify. For those considering careers in AI, Rezkalla recommends specializing in either building models or maintaining the data infrastructure that supports them. "You're always going to need somebody to actually wrangle the data," he says. "There's a lot of technology under the hood that's still being built and developed, and we have to understand how people are using it," he says. "That's what gives us insight into, 'Did we nail the right product experience?'" (cont. next page) "You're always going to need somebody to actually wrangle the data."

• Cover