Wiktor Polonski runs a design studio, an AI consultancy and a GPU infrastructure venture across three countries. Each one grew from the last thing he taught himself.

Wiktor Polonski writes production code in C++, builds WebGL shader pipelines, trains custom AI models for enterprise clients and is now designing a GPU datacenter where investors own physical hardware. He runs three companies across Poland, Spain and Florida, each one a product of the last skill he taught himself.
The path from corporate org charts to GPU infrastructure was not planned. Polonski started in data visualisation, building organisational charts for corporate clients. The work paid well enough to create a financial cushion, and he used it to leave. He moved into game development, studying Unreal Engine and Unity, then taught himself the programming languages underneath: HTML, CSS, JavaScript, Three.js, Python, C++ and C#. Each discipline opened the next project.
That technical foundation became RRAW, a digital atelier Polonski founded in 2019. The studio builds immersive WebGL experiences, brand identity systems, product engineering for SaaS platforms and interactive 3D showcases.
AI-generated design tools have made it easier to produce competent-looking websites and interfaces, which means more of them look the same. Most agencies pull from identical template libraries and component systems. RRAW’s differentiator is that Polonski and his team write custom GLSL shaders and build bespoke Three.js environments from scratch, work that requires the kind of graphic engine knowledge he picked up during his game development years.
‘In the AI-gen era there is a lack of good design,’ Polonski says. ‘Everyone is using the same templates and libraries. This is where we step in and do something that stands out.’
Qubiflow, Polonski’s second company, operates on the other side of the AI equation. Rather than competing against AI-generated design, Qubiflow helps organisations adopt it. The company offers discovery calls, AI coaching for founders, workshops for leadership teams, custom model training on proprietary data and fully autonomous agentic AI assistants that handle customer triage and workflow automation.
The enterprise products run on-premise or at the edge, which matters for clients in regulated industries where data cannot leave their infrastructure. Qubiflow is registered in South Florida and serves clients across finance, healthcare and technology.
Subscribe to our newsletter and never miss a story. No spam, ever.

PayPal still processes $1.8 trillion a year, but the company that built online trust has spent a decade destroying it on both sides of the transaction.

Regulatory complexity and the largest wealth transfer in history are pushing high-net-worth families away from solo advisors and toward coordinated specialist teams.

A forgotten mine in Cornwall's Tamar Valley has just been valued at over £1 billion after drilling revealed far more tungsten, tin and silver than anyone expected. With China choking off 80% of global tungsten supply, the timing could not be sharper.
The connection between RRAW and Qubiflow is Polonski’s own working pattern. He adopted AI tools before the current wave of large language models, using machine learning in design workflows and product development. Qubiflow productises that experience: teaching others the integration techniques he worked out through building.
QUBI-BLOCK is Polonski’s most ambitious project and the one furthest from launch. The concept applies a property investment model to GPU datacenter infrastructure. Instead of renting compute from a cloud provider, investors purchase physical clusters of GPUs housed in a shared datacenter facility and receive revenue from the compute capacity those GPUs generate.
The demand for that ownership route is not abstract. Global datacenter GPU investment reached $21.6 billion in 2025 and is projected to exceed $265 billion by 2035, according to Future Market Insights. Most of that capital flows through hyperscale providers (AWS, Google Cloud, Microsoft Azure) or large GPU cloud companies like CoreWeave, which has raised billions in debt secured against its Nvidia hardware. Smaller investors and businesses have limited options for direct ownership of AI compute infrastructure.
QUBI-BLOCK’s fractional ownership model would allow individual investors to own specific GPU clusters within a managed facility, similar to how a property investor might own a unit in a serviced apartment building. The investor holds the asset; the operator manages the facility, power, cooling and networking. Revenue is shared based on compute utilisation.
The concept is pre-launch, and Polonski has not disclosed the facility location, GPU specifications or financial terms. The timeline is flexible, with announcements expected across 2026.
Across all three companies, Polonski applies the same filter to client selection. He describes the hardest part of running a services business not as finding clients but as finding the right ones.
‘You can have 100 active clients and fail, and have one good one and succeed,’ he says. A good client, in his definition, is ‘someone who has the courage to pursue their dream project and is willing to sacrifice something for it to succeed.’ He frames the relationship as a partnership rather than a transaction, which explains why RRAW offers retainer partnerships that include quarterly strategy sessions rather than just deliverable queues.
Beyond client selection, Polonski wants to build an environment where creators are evaluated on portfolios and ideas rather than CVs and credentials. ‘I believe everyone deserves a chance to show their skills and projects,’ he says. ‘I would like to see more people playing with open cards instead of playing business simulator games.’