Maya: An AI-Native Operating System Kernel

Read full text Summarise with VIRAT
Authors: Jyothi Sarath Kolaparthi
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

We present Maya, an AI-native operating system kernel built from scratch in Rust targeting x86-64 bare metal. Maya departs from conventional OS design by embedding machine learning models directly into core kernel subsystems: a PPO-trained neural network replaces the traditional process scheduler, an AI anomaly detector gates every I/O operation, and a natural language interface allows users to query and reason about kernel state in plain English. Maya is not a simulation or prototype — it boots from USB on real x86-64 hardware in under 2 seconds, detects hardware topology via ACPI, and provides an interactive shell backed by a 3-billion-parameter language model running locally. We evaluate Maya on a Dell Inspiron (Intel Core, 8 logical cores, 16GB RAM) and report IPC latency of 109ns (1.5-4.9x faster than Linux, macOS, and Windows), I/O mediation decisions at 550ns per operation providing continuous AI-driven anomaly detection, and AI scheduling decisions at 41,021 cycles with 0.14% CPU overhead. The kernel passes adversarial security test suites with 13/13 tests passing. Maya's implementation in safe Rust eliminates entire classes of kernel vulnerabilities present in existing C-based kernels. We argue that the overhead of AI-driven kernel decisions is acceptable today and will become negligible as hardware improves, while the security and adaptability benefits are durable.

Fields of Study & Subjects

Parallel Computing and Optimization Techniques Real-Time Systems Scheduling Security and Verification in Computing

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 March 2026
Language
EN