awesome-open-data-centric-ai

Why we included this project

When your model's performance stalls because the training data is messy, this curated index points to open-source tools that fix the data side of the problem. It covers data versioning, embeddings, visualization, and outlier or noise detection for unstructured inputs like images, audio, and text, so you can find a library without digging through dozens of repos. The maintainers keep the scope tight by leaving out tabular tooling, dedicated labeling tools, MLOps platforms, and research papers, which avoids overlap with other awesome lists. A companion playbook collects workflow snippets that show how typical tasks get solved with the listed tools, which helps when you want a concrete example before committing to a stack. It's a discovery resource for teams that want to systematically improve training datasets, not something you deploy directly.