Case Study: Cromai achieves 15x faster deep learning training with DNX Solutions

15x Faster Deep Learning Training in the Cloud

Cromai, a Brazilian AgTech company, uses AI and computer vision to help farmers increase efficiency and sustainability. Their business was being held back by extremely slow deep learning model training, which took approximately three months per model on their existing single-GPU server. This delay directly impacted their core operations, so they partnered with DNX Solutions to find a cloud-based solution.

DNX Solutions migrated Cromai's training to AWS and utilized Amazon SageMaker distributed and the Horovod framework to enable parallel training across multiple powerful GPUs. This solution slashed the model training time from 3 months down to just 6 days while maintaining performance. The drastic time reduction provided Cromai with increased agility and allowed its team to focus on developing improved solutions for agricultural producers.

Cromai