Scale AI workloads with confidence
Anyscale is a scalable AI platform designed for deploying and fine-tuning open-source large language models (LLMs) using Ray-based distributed computing. Developed by the creators of Ray, Anyscale enables users to efficiently run and scale various machine learning and AI workloads, encompassing data processing, training, and inference. The platform supports a range of functionalities, from data preparation to online serving, and is optimized for Python applications. Key features include the ability to fine-tune LLMs for batch inference and entity recognition, utilize reinforcement learning with SkyRL, and leverage Ray's core AI libraries such as Ray Data and Ray Train for large-scale training. Anyscale also offers a fully managed compute service in collaboration with Microsoft on Azure, ensuring users can scale their AI workloads seamlessly while benefiting from hands-on support for troubleshooting and optimization.