TheStage AI Raises $4.5M to Automate Neural Network Optimization and Acceleration

TheStage AI has raised $4.5 million to launch its automated AI model optimization service, designed to significantly accelerate and simplify neural network deployment. The round included investors Mehreen Malik, Dominic Williams (DFINITY), Atlantic Labs (SoundCloud), Nick Davidov (DVC), and AAL VC, along with the Liberman brothers as advisors.

Founded by former Huawei engineers and PhD mathematicians Kirill Solodskih, Azim Kurbanov, Ruslan Aydarkhanov, and Max Petriev, TheStage AI tackles the inefficient manual optimization process traditionally used by AI engineers. The company's flagship product, Automatic NNs Analyzer (ANNA), employs advanced discrete math algorithms and AI techniques, such as quantization, sparsification, and pruning, to automate the optimization of PyTorch neural networks. ANNA reduces optimization time from months to hours, improving performance and cutting deployment costs by up to 5x.

TheStage AI offers subscription-based access to its "Elastic Models," which users can adjust via a single slider to balance between model size, latency, and performance. Its Model Library features optimized versions of popular open-source solutions, including Stable Diffusion. TheStage AI supports diverse hardware configurations, from smartphones and custom GPUs to major cloud providers such as AWS, Google Cloud, and Microsoft Azure.

Customers like Recraft.ai and Praktika.ai have already implemented the technology. Collaboration with Recraft.ai resulted in doubled performance and a 20% reduction in processing time compared to PyTorch’s compiler.

The founding team's research background includes over 10 patents and more than 15 publications, with recognition at leading conferences such as CVPR and ECCV. At Huawei, their work on neural network compression and acceleration was deployed in the AI camera systems of the P50 and P60 smartphones.

According to CEO Kirill Solodskih, TheStage AI simplifies deployment. "We've created a service that allows AI engineers to compress, package, and deploy models to any device as easily as copy and paste."

"We're helping the team demonstrate significant product growth over the coming months," commented Mehreen Malik.

Industry trends underscore the need for TheStage AI’s solution. McKinsey highlights that GPU infrastructure represents up to 70% of AI deployment costs, emphasizing the critical role of efficient optimization, while Deloitte reports that 74% of enterprises have already met or exceeded their generative AI goals.