Determined AI

Determined AI, founded in 2017, provides an open-source deep learning training platform designed to simplify and accelerate the process of building artificial intelligence models. The company was established in San Francisco by Evan Sparks, Neil Conway, and Ameet Talwalkar, a team combining expertise in machine learning and distributed systems. A significant milestone in the company's history was its acquisition by Hewlett Packard Enterprise (HPE) in June 2021, a move intended to integrate Determined AI's software with HPE's high-performance computing (HPC) offerings. Following the acquisition, HPE committed to continuing the development of the platform as an open-source project.

The business targets machine learning engineers and data scientists, helping them to focus on model development rather than managing complex infrastructure. Its platform operates in the machine learning and deep learning market, serving clients across various industries including biopharmaceuticals, autonomous vehicles, and defense contracting. The core of the business is its open-source software, which suggests a business model likely centered around enterprise-level support, managed services, and integration with hardware solutions, particularly following the HPE acquisition.

The platform itself is an integrated environment that streamlines the machine learning workflow. Key features include high-performance distributed training, which significantly speeds up model training times without requiring changes to the model's code. For instance, it reduced the training time for a drug discovery model from three days to just three hours. It also offers advanced hyperparameter tuning to find optimal model configurations, efficient GPU scheduling to maximize resource utilization, and comprehensive experiment tracking to ensure reproducibility. The system is compatible with popular deep learning libraries like TensorFlow and PyTorch and can be deployed on-premises or in the cloud.

Keywords: deep learning training, open source machine learning, hyperparameter tuning, distributed training, GPU scheduling, model management, MLOps, AI infrastructure, Hewlett Packard Enterprise, experiment tracking, neural architecture search, PyTorch, TensorFlow, resource management, AI model development, HPC for AI, enterprise AI, machine learning platform, Ameet Talwalkar, Evan Sparks, Neil Conway