Determined AI

Determined AI, founded in 2017 by Neil Conway, Evan Sparks, and Ameet Talwalkar, provides an open-source deep learning training platform to accelerate the development of artificial intelligence models. In June 2021, the San Francisco-based startup was acquired by Hewlett Packard Enterprise (HPE) with the goal of integrating its software into HPE's high-performance computing (HPC) solutions. HPE has stated its commitment to continue developing the platform as an open-source project.

The company's platform is designed for machine learning engineers and data scientists, allowing them to concentrate on building models rather than managing the underlying infrastructure. It serves a diverse client base across industries such as biopharmaceuticals, autonomous vehicles, and manufacturing. The business model, especially after the HPE acquisition, is centered around its open-source software, likely generating revenue through enterprise support, managed services, and hardware integration.

The platform offers an integrated environment that includes high-performance distributed training, which can significantly reduce model training times—in one drug discovery case, from three days to three hours. Key features include advanced hyperparameter tuning for optimizing model configurations, efficient GPU scheduling, and comprehensive experiment tracking for reproducibility. The system is compatible with popular deep learning libraries like TensorFlow and PyTorch and supports both on-premises and cloud deployments.

Keywords: open-source, deep learning, training platform, machine learning, MLOps, distributed training, hyperparameter tuning, GPU scheduling, model management, AI development, high-performance computing, experiment tracking, reproducibility, TensorFlow, PyTorch, developer tools, AI infrastructure, neural architecture search