Graviti

Founded 2019

Curate, version, and manage unstructured datasets for machine learning and analytics workflows

Pricing model

Free Tier, Paid

You can use Graviti to store, curate, and version unstructured data for machine learning models. It lets your team manage raw files, metadata, and annotations in a single location with Git-style version control and branching. You can inspect data quality, identify imbalanced or low-quality data points, and query samples without duplicating files. Graviti also allows you to automate data preprocessing steps like augmentation and auto-labeling, visualize differences between dataset versions, and trigger training pipelines whenever fresh data arrives.

What Graviti does

Create branches to test dataset modifications safelyCompare visual differences between two dataset versionsFilter raw datasets to generate new training subsetsAutomate preprocessing and labeling pipelinesInspect metadata and label accuracy across filesDetect under-represented classes in training setsGit-like data version control with branching and commit historyZero-copy dataset curation and filteringVisual comparison interface to view differences across dataset versionsAutomated data preprocessing pipelines for augmentation and auto-labelingData quality inspection to find incorrect semantic labels and metadataDetection tools for identifying imbalanced and under-represented dataAutomated model training triggers when new data batches are added

Pricing breakdown

Pricing model: Free tier, Paid

PlanList price

StarterFree

Standard$200 / mo

Premium$800 / mo

Frequently asked

MLflow, Git, MLflow for experiment tracking, Amazon S3 for storage, Google Cloud Storage, Azure Blob Storage, Python SDK, PyTorch data loaders, TensorFlow data loaders

Want a tailored answer?