At AI ATOM GLOBAL EXPO 2026, NeuroDrift, a project that trains a sports car to drift under the control of a neural network, signed a partnership agreement with Alan Kidson. The partner will be responsible for commercializing the project in the United States, while NeuroDrift will focus on developing and supporting the technology.
According to project lead Konstantin Nikiforov, this is NeuroDrift’s first agreement involving mutual commitments with a foreign partner and its first step toward real operations on another continent. A legal entity has been established in the US for the project, and the partner is set to be fully integrated into the company’s development.
“We have many partners and investors who closely follow the results of our teams. This agreement sends a signal to all of them: we are ready to move from partnership to full-scale collaboration. For the first time, the commitments are mutual: our partner is responsible for commercializing the project in the US, and we are responsible for technology development and the technical side. Our relationship is reaching a new level, and we are now starting real work on two continents. I am confident that our other partners will be able to follow the same path”, – said Konstantin.
Konstantin noted that signing an agreement of this kind is a major achievement for Belarus’s startup sector and a signal that young, ambitious projects developing cutting-edge technologies can attract Western investors and enter into full-fledged cooperation.
NeuroDrift is built on the hypothesis that a sports car can be taught to perform the most complex drifting maneuvers using a neural network rather than a controller or a mathematical model. The company has built its own drift car and the entire neural network training pipeline, from data virtualization to transferring skills to the real world. The team’s main achievement is narrowing the so-called sim-to-real gap, the difference between a virtual twin and reality. The car’s virtual behavior matches its real-world behavior with a score of 87 out of 100.
“Usually a skill gets lost in the transfer, and that is a problem, but in our case it is preserved: we train the neural network in a virtual environment using a target task and a reward function, and then transfer the acquired skill to a real car”, – Konstantin added.
The project’s ultimate goal is to launch its own world championship league, where racing drivers and technology companies with their own AI models can compete against each other. The company estimates the potential audience for this format at around 2.1 billion people, the total audience interested in motorsport.
Beyond its entertainment and marketing potential for car, tire and lubricant manufacturers, the project also has engineering applications. The pipeline it has developed can be adapted to automate any type of vehicle, including:
- agricultural machinery;
- space exploration vehicles, such as planetary rovers;
- deep-sea vehicles that lay fiber-optic cables on the ocean floor, and more.
About NeuroDrift
NeuroDrift is a technology project in the field of autonomous motorsport and machine learning. Its team teaches a neural network to drive a real sports car and perform controlled drifts without remote control or a predefined trajectory. To achieve this, the team has built its own drift car and a unique process that allows the neural network to learn in a virtual environment and then transfer the skill to the real world without loss.
About AI ATOM GLOBAL EXPO 2026
The international forum on artificial intelligence and digital technologies took place in Minsk on October 16-18. Across 20,000 m² of exhibition space, it featured more than 50 AI projects and startups and over 30 speakers from Belarus, China and Europe. Visitors could try the NeuroDrift technology for themselves at the stand: take a seat in a racing simulator and compare their result with that of the neural network. Learn more at atomexpo.by.
