![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 60 45" xmlns="http://www.w3.org/2000/svg"><path d="M 22.5 22.5 L 37.5 45 L 15 45 L 0 22.5 L 15 0 L 37.5 0 Z M 4.049 22.5 L 16.803 41.631 L 31.204 41.631 L 18.451 22.5 L 31.204 3.369 L 16.803 3.369 Z" fill="rgb(255, 255, 255)" height="44.9999542236px" id="D2n7KwITb" transform="translate(22.5 0)" width="37.5px"/><path d="M 0 22.5 L 22.5 22.5 L 22.5 0 Z M 19.191 19.191 L 7.988 19.191 L 19.191 7.988 Z" fill="rgb(255, 234, 0)" height="22.5px" id="r8yrV6Ym6" transform="translate(0 22.5)" width="22.499936117px"/><path d="M 0 22.5 L 0 0 L 22.5 0 L 22.5 22.5 Z M 19.191 3.309 L 3.309 3.309 L 3.309 19.191 L 19.191 19.191 Z" fill="rgb(255, 0, 26)" height="22.50000205642px" id="HkttCdpAI" transform="translate(0 0)" width="22.5000134408px"/><path d="M 10.227 5.114 C 10.227 7.938 7.938 10.227 5.114 10.227 C 2.289 10.227 0 7.938 0 5.114 C 0 2.289 2.289 0 5.114 0 C 7.938 0 10.227 2.289 10.227 5.114 Z" fill="rgb(255, 0, 26)" height="10.227310000000001px" id="X_Hp2kGsw" transform="translate(6.257 6.396)" width="10.227339999999998px"/><path d="M 6.617 5.113 C 6.617 4.283 5.944 3.609 5.113 3.609 C 4.283 3.61 3.609 4.283 3.609 5.113 C 3.609 5.944 4.283 6.617 5.113 6.617 L 5.113 10.227 L 4.85 10.221 C 2.149 10.084 0 7.849 0 5.113 C 0 2.289 2.289 0 5.113 0 C 7.937 0 10.227 2.289 10.227 5.113 L 10.22 5.377 C 10.083 8.079 7.849 10.227 5.113 10.227 L 5.113 6.617 C 5.944 6.617 6.617 5.944 6.617 5.113 Z" fill="rgb(255, 0, 26)" height="10.227310000000001px" id="amWlbbgtf" transform="translate(6.257 6.396)" width="10.227339999999998px"/></svg>)
Nebius
Nebius announced Token Factory, their new production-scale inference platform built for high-throughput, low-latency AI. My role was to turn that technical leap into a visual universe — something fast, expressive, and unmistakably “open-source-powered.”
Built for Scale
The Token Factory announcement came with a clear point of view: turn open-source models into production tools. That idea shaped the visual logic. The halftone textures acted like a signal — data in motion. I treated every part of the brand as a working component.
Engineered Motion
Even with a very modular layout, there was room for personality. Small distortions, subtle scattering, and unexpected shifts in depth gave the system a bit of life without drifting off-brand.
Tokens were rendered as clean geometric objects — simple, repeatable, and easy to deploy across scenes. Inference paths stacked and curved to build rhythm. Node clusters created a clear visual bridge to governed collaboration and post-training. Everything had a job, and every job tied back to the core product story.
A System That Works Hard
The final film introduced Token Factory with focus and momentum. It framed the platform not as another AI tool, but as the infrastructure layer where inference, post-training, and governance meet. The motion proved that a technical product can still feel expressive when the system is built with intention — fast when it needs to be, restrained when it matters, always grounded in clarity.
The piece became the starting point for how Nebius can show its technology across product, marketing, and future launch campaigns, setting the tone for a brand that moves at production speed.
