That’s why German IT service provider isento built the TRYB platform: to make AI accessible to everyone. TRYB is a collaborative, no-code solution that enables organizations such as manufacturing companies and researchers to go from an idea to a trained model in minutes. Through a simple interface, users can easily train AI for tasks like visual quality assurance or programming industrial robots, allowing them to bypass technical complexity and focus on making the most of AI.

As isento set about building the TRYB platform, it needed a cloud partner that would be simple to use and offer integrated tools to save development time. In particular, the company needed the capability to manage and rapidly duplicate massive image databases as part of its AI model development process.

“It was clear we’d be able to build the features we needed quickly and easily with Google Cloud,” explains Shota Okujava, CEO and co-founder of isento. “The native functionality of BigQuery gave us a way to rapidly clone large image databases that would have been impossible with other platforms.”

To innovate effectively, TRYB users need to be able to rapidly test and iterate AI models. For visual AI tasks, this means duplicating and modifying massive training image datasets, which traditionally leads to prohibitive storage costs and slow workflows. Using the zero-copy cloning capability of BigQuery, the platform instantly creates a new, independent copy of a training dataset by cloning the metadata, without duplicating the underlying data. This keeps the data consistent, while accelerating the process and controlling costs.

Cloning an entire image database of more than 30,000 images takes less than 10 seconds, allowing users to immediately iterate and try out new ideas. And by avoiding data duplication, an average project with multiple training iterations requires about 85% less storage space than the traditional full duplication on a third party provider. This equates to a similar reduction in storage costs, helping Isento to keep the price of its services as low as possible.

With Vertex AI, meanwhile, TRYB enables users to train sophisticated models for complex tasks. This could include object recognition for quality control or developing motor skills for a robotic arm in a “digital twin” environment, which reduces wear and tear on expensive physical robots, extending their lifespan.

Once an AI model is successfully trained, isento uses Artifact Registry to store the containerized models and Cloud Run for deployment. The scalable, serverless API endpoint of Cloud Run — which is crucial for delivering a cost-effective platform — means the framework automatically scales resources based on user demand, from a single user to thousands. This allows isento to offer a flexible, pay-as-you-go model, further lowering the barrier to entry for businesses.