In this presentation, we will introduce a machine learning platform developed in Rust, specialized for real-time applications.

There are various systems that deliver machine learning inference results to users. In many cases, models are pre-trained offline, and during inference, the pre-trained models are loaded to set up an API. However, in scenarios such as reinforcement learning, bandits, and services with rapidly changing content, there is a demand for conducting training online as well. The machine learning platform we are introducing addresses these needs by enabling not only offline training but also real-time online training. With Rust gaining attention and expanding its applications, we hope this serves as a reference for how it is being applied in real-world machine learning services.