hyperparam

hyperparam is the missing UI for machine learning

Building the best model requires building the best dataset.

The first step in data science is to be deeply familiar with your training data. But where do you even start? Modern LLMs depend on terabytes of unstructured text data. Most data tools cannot handle this scale of data interactively, or require sampling to show only a tiny slice of your data. Hyperparam brings visualization and analysis of massive text datasets to the browser.

Everyone evaluates models, we evaluate datasets.

To find good quality data, you need to evaluate your dataset. Furthermore, data quality is application specific – the real question is: how do you find the good training examples for your use case, and remove irrelevant examples. To find these at scale, we use machine learning models to reflect back on their own training set.

Building the best dataset requires faster iteration.

Speed of iteration is the key to success in machine learning. Hyperparam combines automated methods for shaping your dataset with a scalable UI for human-in-the-loop feedback. This enables a powerful iteration cycle: collect data, train a model, use that model to refine its training set, and repeat the process until you achieve the highest-quality model.