ZeroML: A Next Generation AutoML Language

TL;DR

ZeroML addresses the slow runtimes, brittle pipelines, and dependency fragility of existing AutoML tools built on Python, R, and Julia by introducing a compiled, multi-paradigm language with a pure functional core and modular AutoML components. The approach combines procedural and object-oriented styles, enabling modular reuse of components like DataCleaner, FeatureEngineer, and ModelSelector while supporting memory-efficient, multithreaded execution. It contrasts with current stacks by offering concise syntax, reduced boilerplate, and one-line deployment while scaling from local machines to distributed environments. The work promises faster prototyping, reproducible workflows, and broader accessibility for non-coders and experts alike, ultimately enhancing deployment readiness and efficiency in AutoML pipelines.

Abstract

ZeroML is a new generation programming language for AutoML to drive the ML pipeline in a compiled and multi-paradigm way, with a pure functional core. Meeting the shortcomings introduced by Python, R, or Julia such as slow-running time, brittle pipelines or high dependency cost ZeroML brings the Microservices-based architecture adding the modular, reusable pieces such as DataCleaner, FeatureEngineer or ModelSelector. As a native multithread and memory-aware search optimized toolkit, and with one command deployability ability, ZeroML ensures non-coders and ML professionals to create high-accuracy models super fast and in a more reproducible way. The verbosity of the language ensures that when it comes to dropping into the backend, the code we will be creating is extremely clear but the level of repetition and boilerplate required when developing on the front end is now removed.

Figures (4)

• Figure 1: Overall Structure of proposed ZeroML Language.
• Figure 2: Code snippet with Python Language (TPOT).
• Figure 3: Code snippet with R Language (H2O).
• Figure 4: Code snippet with ZeroML Language.