Jason Rieff once reported a profit of $84.87 from an experiment that began with wall stickers for his son's room. He had looked at decals online, disliked the prices, and decided to make his own with image-generation tools. Then the marketer in him took over. Why stop at decorating a bedroom when there was an advertising platform to investigate?
He made the images, arranged printing and opened an Etsy shop. His first attempt at paid advertising failed. He watched tutorials, tried again and added QR-code stickers around his neighbourhood. A station lift proved a useful location: parents with pushchairs needed it. Seven sales later, his little test had produced revenue of $161.82 against spending of $76.95.
The sums are charmingly small beside the enterprise AI business he now runs. His explanation is more revealing than the profit. “In reality, I paid myself by learning,” he wrote. That is a useful place to meet the co-founder and co-CEO of BeatpulseLabs: in the middle of an experiment, checking what actually works.
The guest was never just an age bracket
Before AI training data, Rieff worked on a different classification problem. As chief marketing officer at Generator, he helped sell a hospitality business whose offer stretched from shared rooms to private suites. Calling everybody a young backpacker would have been convenient. It would also have missed a fair number of the people walking through the door.
In 2019, he described an audience that could include a suited traveller and an adventurous grandmother. His preferred organising idea was a shared mindset. Music, partnerships and public spaces were chosen around what guests wanted to experience. At the time, Generator had 14 properties, with Washington, DC on the way. The business had to make sense across cities without making every city feel identical.
There was a practical mechanism behind that thinking. Generator followed listening and viewing signals from SoundCloud, YouTube and Spotify to inform music across its properties. Local artists and brands helped shape individual locations. Rieff worked with CEO Alistair Thomann as the business invested in booking technology, marketing and partnerships. The brand work reached well beyond the artwork on the wall.
One small detail conveys his impatience with hospitality by assumption: he objected to the decorative letter cutouts that hotels imagined guests wanted. It is an amusingly precise enemy. Anyone can declare an interest in authenticity; identifying an unnecessary bit of lobby typography takes a closer look.
In October 2018, Generator launched GenFriends, an app through which travellers could meet, exchange recommendations and find offers. Rieff presented it as an extension of the company's social spaces. People were already gathering in the buildings. The app offered another way for them to find one another, including when travelling alone.
Viewed together, these decisions suggest an early version of a question that would follow him into another industry: how much do you lose when you describe people, or their work, with a label too broad to be useful?
“Know your customer and listen to your research.”JASON RIEFF · 2019
A consultancy, then a question about music
Rieff brought an agency background to hospitality. Before Generator, he had been national strategy director for We Are Social in Germany. In 2019, he founded JSR-Consultancy, working with a core group of external specialists on brand building and performance marketing. Its services included search advertising, social media, partnerships and customer relationship management.
The combination matters. Brand strategy asks what a business means to people. Performance marketing asks whether those people do anything about it. Rieff's consultancy put both kinds of work on the same menu. The small Etsy experiment makes sense in that professional context: a chance to try the tools, buy some traffic and see where the assumptions broke.
By 2021, music rights had become a public focus. As co-founder of Noizd, an NFT music marketplace, he discussed the leverage record companies held over artists who needed income and exposure. His objection concerned the pressure to accept poor terms when the alternative might be obscurity. He saw NFTs as a possible route for artists to earn outside that arrangement.
That was an argument made in a particular technological moment. Its place in his story is the attention to the relationship between creators, ownership and access. His F6S profile described Noizd's ambition in terms of music rights and African creator economies. The proposed technology was new; the question of who gets paid for creative work was considerably older.
A music business learns another language
Rieff subsequently co-founded Beatpulse. In a public announcement, he celebrated its admission to the first Techstars Founder Catalyst programme in Jeddah, naming fellow founders Theerasak Charoenchob and Suksan Salarak. The programme connected his music venture with a wider group of digital-media businesses. It was another setting in which licensing, technology and commercial experimentation met.
In July 2024, he appeared with Paul Essery on The Tech Business Podcast to discuss AI and digital music content. Later that year, his Web Summit profile listed him as BeatpulseLabs' CEO and placed the business in generative music and audio training data. It also described him as a three-time founder with one exit.
Those appearances mark a change in the language surrounding his work. The hotel marketer had talked about audiences and experiences. The music entrepreneur talked about rights. At BeatpulseLabs, the vocabulary included datasets and model training. Read as a career, the movement is less bewildering than the changing nouns suggest: each business puts a different commercial problem between creative material and the people trying to use it.
The work inside the label
Today Rieff shares leadership of BeatpulseLabs with co-founder and co-CEO Nikolay Vitanov. The company builds multimedia training datasets for enterprise AI. Its stated approach centres on specialists, annotation designed for individual clients, and feedback from the people doing the work. A dataset is treated as something that must fit a particular use.
Its company account of the problem is straightforward. Broad public datasets and general-purpose labels can leave out the distinctions an enterprise needs. BeatpulseLabs says it develops annotation schemes with those requirements in mind, using domain specialists in structured teams. That is its commercial proposition, rather than a guarantee about every model trained on its material.
- 01Raw materialMusic, speech and video
- 02Expert judgementLabels built for the task
- 03ValidationAmbiguity and quality checks
- 04Training dataStructured for the model
The music offering makes the idea tangible. Its annotation work includes descriptions of tracks, interpretation of genre and emotional tone, and identification of compositional sections. The company also offers recordings separated into stems, alongside MIDI and other musical assets. A label such as “music” plainly leaves a lot of work unfinished.
Speech introduces another set of distinctions: who is speaking, when a segment begins, whether a speaker changes languages, and how tone or intent should be described. BeatpulseLabs advertises transcription, accent tagging and these less literal layers of speech annotation. Video requires attention to shots, scenes, subjects and the alignment of picture with sound. Each medium brings its own opportunities for a careless label.
There is a pleasing complication in Rieff's current occupation. A business helping machines learn sells the value of human judgement. BeatpulseLabs describes specialists flagging ambiguity and difficult examples, then feeding that information back into the annotation process. In that account, expertise continues to shape the material while the system is being built.
A larger experiment, with other people involved
Rieff has publicly said BeatpulseLabs supplied part of the training data for ElevenLabs' music model. He has also announced data-provider relationships with Korus and Beatoven.ai. These are specific accounts of where the company's work has been used, and they give the abstract language of AI infrastructure a recognisable musical destination.
His posts also show the importance he places on collaborators. When acknowledging investors, he singled out Vitanov as both his co-founder and friend. Their shared title places two people at the top of the organisation. It would be tidy to assign one the technology and the other the business, but the public titles themselves do not make that division.
Co-led by Arāya Ventures and Lighthouse Ventures, with Alumni Ventures and Avalancha Ventures participating.
The June 2026 financing was accompanied by company-reported tenfold revenue growth during the first half of the year. The announced purpose was expansion of the platform and customer base. For Rieff, that sets a different scale of test from the neighbourhood sticker shop: repeatable delivery, enterprise requirements and a business built with a co-founder, employees and investors.
The useful thread through these episodes is his attention to the work between an idea and a usable result. A guest has to feel at home. A musician needs a workable route to market. A model developer needs material organised for a particular task. The details change; the obligation to understand the user survives.
The decal story supplies a modest ending for a career now surrounded by large numbers. He tried a tool, spent money, got something wrong and adjusted. The profit was $84.87. The more interesting return was an answer he could use.