Hewlett Packard Enterprise
Poor quality ML data is the biggest impediment to fast productionization of ML models — yet, managing data across training and production models is ad-hoc, time consuming, highly manual and error-prone today, especially for unstructured data. This leads to reactive training data fixes for production models, biases creeping in and poor model predictions.
Galileo is a first of it’s kind ML Data Management tool to enable quick error (mislabels, poor predictions, etc) analysis, production<>training drift detection, smart training data sampling from production data and data/model tracking – all in one, collaborative space – ML teams, SMEs, PMs, Labeling teams all use Galileo to enable data transparency and high quality training data that is always representative of the real world.
We are well funded, built on the founding team’s past work leading product and engineering at Google AI and Uber AI, and are in private beta working with ML teams at F500 and fast moving early stage enterprises that have critical ML models using unstructured data, starting with NLP.