Universal Music Group’s latest AI hire built his career on recommending music.
On Sept. 28, UMG named Òscar Celma its Senior Vice President of Applied AI and Machine Learning, the first hire of its kind at the biggest label group.
Most coverage of the hire focused on Spotify, where Celma helped lead development of AI DJ, Jam and Daylist. The rest of his career fills in the picture. He spent seven years in data science leadership at Pandora/SiriusXM, worked as a research scientist at Gracenote, co-founded the music-identification company BMAT, and literally wrote a Springer book on music recommendation systems.
For two decades, his work has focused on two questions: what should someone hear next, and what exactly is playing?
Those are useful skills for the world’s largest music company to bring in-house right now. As UMG works to shape how AI music is made, licensed and distributed, it needs to know whose work is inside every output and how that music gets surfaced.
Celma is just one hire at one company. But he is the latest data point in a much broader pattern playing out across the music business, showing how both labels and AI music companies are staffing for what comes next.
Labels are recruiting around data, rights and technology. AI music companies are importing licensing, artist-relations and record-business expertise. And the jobs still open show both sides preparing for the same challenge: turning generative music from a model into a functioning music business.
Now that a model can generate a song, the companies need people to license it, credit it, surface it, and pay for it.
Who the labels have hired
UMG is rapidly building data and AI infrastructure. Celma reports to Hannah Poferl, who joined as Chief Data Officer in January after serving as The New York Times’ first CDO. At the Times, she used data to grow and engage a paying subscriber audience — which is the kind of relationship UMG wants with fans. COO Boyd Muir tied her hire to UMG’s superfan strategy and direct-to-consumer platforms.
Its UK arm is recruiting AI engineers for A&R insights and rights and royalty workflows, including automating royalty calculation.
Those hires are happening alongside UMG’s growing AI licensing business. A growing part of its superfan push runs through AI: it has licensed fan-creation and remix platforms to Udio, ElevenLabs and Spotify, and every one of those systems will produce outputs that need to be identified, credited and paid.
Warner Music Group promoted Leho Nigul to Chief Technology Officer in late 2025, days before it began settling lawsuits and turning AI companies into licensees.
Nigul’s background is in the plumbing. Before the promotion, he led Warner’s engineering on global licensing, royalty management, and AI projects. Since then, CEO Robert Kyncl has told investors that AI licensing should contribute materially to subscription streaming revenue starting in fiscal 2027.
Warner put the person who knows how licenses become royalty statements in charge of all of its technology, right as AI licensing became a line in its revenue forecast.
Sony Music hired for a different job. In 2023 it created the first AI-titled executive role at a major label, EVP of Artificial Intelligence, and filled it with Geoff Taylor, a lawyer and longtime anti-piracy leader who had spent 15 years running the British Phonographic Industry.
Sony’s strategy resembles his résumé. A year after his hire, it sent letters to roughly 700 AI developers and streaming services indicating it was opting out of AI training with its catalog. The major still has active litigation against Suno and Udio, while also licensing companies such as Klay Vision on its own terms.
UMG and Warner put builders in charge. Sony hired someone who spent his career deciding who gets to use music, and on what terms.
None of the three majors is building its own AI generation model at scale. When they want exposure to that layer, they invest in it or license it. UMG, Sony, and Warner have all invested in Stability AI, while keeping the catalog, data and payment systems closer to home.
Who the AI music companies have hired
The AI companies are hiring the mirror image of what the labels are.
Suno is the strongest evidence. It has the dominant product in the generative AI music space and a $5.4 billion valuation. What it lacks is licenses from UMG and Sony, which have active lawsuits against it. So it has spent the past year trying to bolster its knowledge and credibility by hiring from within the record business.
It brought in Paul Sinclair, Atlantic Records’ former general manager, as Chief Music Officer. Sam Berger joined to run artist partnerships after two years as Patreon’s Head of Music Partnerships, where he built its strategy for bringing musicians onto the platform. He’d also worked on Spotify’s artist and label partnerships team.
In February, Jeremy Sirota became Chief Commercial Officer. As CEO of Merlin, he spent years on the other side of the table, licensing independent labels’ music to platforms, and he’d previously worked at Warner Music and on Meta’s music team. Now he runs Suno’s industry relationships and commercial deals.
In July, Suno added Grace James, Atlantic’s former creative marketing chief, whose job is to market artists who come up through Suno, including its Spark incubator. Christian Bowne joined from YouTube, where he ran major-label business development. He now leads Suno’s licensing strategy, which is the job of getting UMG and Sony from suing Suno to licensing it.
ElevenLabs made one key hire. In January, Derek Cournoyer left Kobalt, the independent publisher, after about a decade shaping its AI strategy and its deals with Meta, TikTok, and YouTube. He became ElevenLabs’ strategy lead for music business affairs, managing relationships with labels, publishers, and collecting societies. At Kobalt, he had helped negotiate the deal that lets Kobalt’s songwriters opt into training ElevenLabs’ music model.
Google already had the model builders. DeepMind’s team built the Lyria music model. What it lacked was a consumer generative-music product and a team built around music creators. In February it acquired Producer.ai, the startup that began as Riffusion, and its entire team joined Google Labs and DeepMind. Google brought in the founders — who are musicians as well as engineers — and the product, then swapped out their model for Lyria. The result became Flow Music.
Klay Vision, licensed by all three majors to build a service that lets fans remake and reshape songs, has a leadership team that contains the whole thesis. Co-founder Thomas Hesse was Sony Music’s president of global digital business. Chief AI Officer Björn Winckler led Google DeepMind’s music work. CTO Brian Whitman co-founded The Echo Nest, the music-intelligence company Spotify bought in 2014 to power its recommendations.
In three executives, Klay has assembled the three kinds of expertise the industry is now converging around: music business, model research and recommendation technology.
The open jobs show what’s next
The hires show what companies decided they needed. The open jobs show what they still lack — and where their priorities lie.
As of Oct. 1, Suno’s careers page listed 66 openings. Two directors of publishing partnerships would pursue songwriters’ rights. A senior counsel for music business affairs would negotiate deals. A director of editorial and curation would decide what gets surfaced. Four trust-and-safety roles would police abuse, including voice cloning.
Only three were listed under machine learning.
Suno has its model. Now it’s staffing the music business around it.
ElevenLabs is looking for music leadership rather than a large music division. Of roughly 170 open roles as of Oct. 1, only one is on the ElevenMusic team: a Head of Music to run licensing, product, marketing and partnerships across consumer and enterprise music. The role sits atop a much larger enterprise sales and deployment machine.
Udio needs artists. Its forthcoming licensed platform, built out of the UMG and Warner settlements, depends on artists opting in their voices and compositions. A Head of Artist Partnerships posting to recruit them has been open since January.
While AI companies hire the music business, Spotify is hiring model builders. In May it brought in Julian Parker, a senior researcher on Stability AI’s Stable Audio models. Its AI music lab is now recruiting research scientists for vocal synthesis and editing tools that can swap stems or change a track’s mood and tempo — capabilities behind the forthcoming AI remix product UMG and Merlin have signed on to.
There’s a reason the hiring is spreading beyond the models themselves. Every license and settlement creates obligations.
Opt-in artists need relationships. Generated outputs need attribution. Rights have to be tracked across recordings, compositions and voices. Money has to be split output by output. And as generated supply explodes, recommendation matters more, because someone has to decide which of millions of new tracks anyone hears.
The model has become just one layer of a larger system. The hiring shows where the hard work is moving next: into the people and infrastructure that license the music, bring artists in, identify what was used, decide what gets heard and make sure everyone gets paid.
Making a song is no longer the hard part. The hiring makes that clear.
🆕 New AI Music Releases
Got a new release? Submit it here.
Albums/EPs
Corin & The Unfinished - Mga Awit Ng Buhay Ko (A Florante Magsipoc Collection) (Singer-Songwriter)
Duchess Of The Damned - Hey!
Ómarh - The Unspoken Things (R&B / Soul)
PrettyLittleVandal - Pretty Little Vandal (Rock / Alternative)
xLesserxEvilx - Daughters of Darkness (Rock / Alternative)
Singles
Alex Viegas - “Luntherland” (Electronic / Dance)
Alibi Ivory - “Deepfake It” (Pop)
Ay Music - “Miss Trouble” (Dancehall)
Emma - “Tell Me Who It’s For” (Pop) and “Sing Anyway” (Pop)
Fenrisúlfr - “The Legion” (Rock / Alternative)
fusionbyarif - “Changed, You’re Right” (Pop) and “Again!” (Pop)
Honest Bob - “Take Me Back (remix)” (Electronic / Dance)
Ione - “Not That Kind of Elf” (Pop)
LucKyNobody Music - “Don’t Know What to Do” (Folk)
Noah & Felix - “IN BETWEEN (Challenge Mix)” (Pop)
PantheraSounds - “Taking It Back” (Rock / Alternative)
Reina Acevedo - “Me Acuerdo” (World / Global)
The Gyaru Twins - “Okinowa Stomp” (Electronic / Dance)
Valentyna Volkov - “Contradiction” (Dark Metal)
🎧 Song of the Day: “Through Me” by Solyn & Rizzards
Rizzards x Solyn’s “Through Me” captured Viewer’s Choice honors at Week 39 of the Fresh Friday Festival, channeling the dizzying, sweat-soaked release of shedding self-consciousness in a packed room. Rather than playing to an audience, the video tracks that exact tipping point where the anxiety of being watched dissolves into the electric momentum of a crowd moving as one. Check out Professor Jacket’s exclusive, full Week 39 recap on Zinstrel.com.
💬 Last Word
“I think both can exist. Everyone’s so black and white about it, but both [AI music and traditional music] can be creative in a completely different way.”
— Grammy-winning singer and talk-show host Kelly Clarkson, to the Las Culturistas podcast
MORE ZINSTREL:
Zinstrel.com | @zinstrelculture on IG | AI Underground on Discord
Zinstrel #121: The Secret (and Not So Secret) Identities of AI Artists
Zinstrel #120: AI Labels Now Reach 60% of Streaming Music
Created by Marcus Lawrence
Copyright 2025-2026 Zinstrel, LLC - All rights reserved
Special thanks to our paid subscribers: Blue Hermit, Bukar Gadzama, C.Y. Lee,
Charles S, Chez, Daniel Drolet, Daryl Dekking, Dave, Dexter Garcia, Drew Thurlow,
Ghost Ryder, Itai Leibowitz, Jacci Cenci-McGrody, Lenny Skolnik, Lynn Clapp,
Marc Wolinsky, Matthew Marturano, Matti Kuha, Mete Dibi, Rosanna, Sheila Spence,
Steve Clapp, Steve Grossman, Tim Burley, and Tracy Maddux.