CentML, a startup specializing in optimizing the deployment of machine learning models, announced it secured $27 million in additional funding for its seed round. Notable investors include Gradient Ventures, TR Ventures, Nvidia, and Microsoft Azure AI VP Misha Bilenko. The initial seed round closed in 2022 but was extended due to growing interest, resulting in a total of $30.5 million raised.
The new funds will be utilized to enhance CentML’s product development and research, expand its engineering team and workforce, which currently spans the U.S. and Canada.
Co-founder and CEO Gennady Pekhimenko, an associate professor at the University of Toronto, founded CentML to address challenges in the AI sector. These challenges include rising machine learning costs, talent shortages, and difficulties in sourcing AI chips. Many companies rely on GPU-based hardware for training models, but shortages in these chips have become prevalent.
Microsoft and Nvidia have both faced severe chip shortages. To combat this, some companies are exploring custom chip solutions, although they encounter challenges in their development. CentML aims to make models run more efficiently on existing hardware with its optimization technology, potentially reducing expenses by up to 80% without sacrificing speed or accuracy.
CentML’s software identifies bottlenecks during model training, predicts deployment time and costs, and offers a compiler to optimize model training for target hardware. According to Pekhimenko, this software requires minimal effort to implement and doesn’t degrade model accuracy.
While other companies like MosaicML and OctoML also offer software-based model optimization, CentML claims to maintain model accuracy better and provides a more performant compiler. CentML plans to expand its focus to optimizing model inference in the near future. This could enable teams to deploy models on smaller, more cost-effective GPUs, making the platform versatile for AI and machine learning companies.
