Ludwig AI

Ludwig AI is an open-source, low-code framework for building custom AI models — from classic neural networks to fine-tuned large language models — using simple declarative configuration files instead of boilerplate training code. Originally created at Uber and now stewarded by the Ludwig AI community with Predibase, it remains one of the most approachable ways to train production-grade models. (It is unrelated to the Ludwig subtitle-translation app also listed on Tooloso.)

Overview

With Ludwig, you describe your data and your task in a YAML configuration: which columns are inputs, which are outputs, and what types they are. Ludwig then assembles the right model architecture, training loop, and evaluation pipeline automatically, on top of PyTorch. It handles text, images, tabular data, audio, and multimodal combinations, and supports fine-tuning popular open LLMs such as Llama and Mistral with techniques like LoRA. The project has more than 11,700 GitHub stars, is under active development with updates in October 2026, and is licensed under Apache-2.0.

Key Features

• Declarative YAML configs replace hundreds of lines of training code
• Fine-tune and train LLMs, plus models for text, image, tabular, and audio data
• Automatic architecture assembly, training, evaluation, and model export
• Built-in support for LoRA and other parameter-efficient fine-tuning
• Hyperparameter optimization and rich experiment tracking
• Serve trained models with a REST API or export for deployment

💳 Pricing

Ludwig AI is completely free and open source under the Apache-2.0 license. You only need your own compute for training and serving.

Who Is It For

Data scientists, ML engineers, and developers who want custom models — especially fine-tuned LLMs — without writing and maintaining training infrastructure.