New Proposal Explores Machine Learning Assistance for Linux Kernel Behavior
New Proposal Explores Machine Learning Assistance for Linux Kernel Behavior
https://linuxiac.com/new-proposal-explores-machine-learning-assistance-for-linux-kernel-behavior/
Publish Date: 2026-02-07 14:48:00
Source Domain: linuxiac.com
A new discussion on the Linux kernel mailing list is exploring whether machine learning could assist kernel subsystems. The idea was proposed by Viacheslav Dubeyko, an IMB engineer, to create a generic infrastructure that allows kernel subsystems to interact with machine-learning models running entirely in user space.
At the core of the proposal is a lightweight “ML proxy” inside the kernel. This proxy would expose structured data from a kernel subsystem, such as internal state or performance metrics, and receive recommendations generated by a user-space ML model. Training, model execution, and experimentation would remain outside the kernel, while the kernel retains full control over whether and how recommendations are applied.
And since ML immediately brings to mind AI, I want to make the following point clear. The idea does not involve embedding AI or ML code inside the kernel. Instead, kernel subsystems would expose data to the user space, where machine-learning models normally run, and receive optional recommendations in return.
In other words, the kernel would remain fully deterministic and in control, deciding at any time whether to apply, test, or ignore those suggestions, with all training and inference kept outside kernel space. Existing mechanisms such as sysfs, character devices, FUSE, or eBPF are cited as possible transport and execution layers.
Dubeyko also outlines a feedback loop in which the kernel evaluates applied recommendations and reports efficiency metrics back to the user space. According to him, this information could be used to refine or retrain models without impacting kernel performance or determinism. Importantly, the proposal emphasizes that human-written kernel logic remains the baseline, with ML acting only as an advisory layer.
An early proof-of-concept implementation of the proposed ML library has been published, and an RFC patch series has been posted to the Linux kernel mailing list for discussion. The…
