What is Hyperparam?

Hyperparam enables exploration and curation of massive ML datasets. By combining 1) a highly scalable browser-based UI with 2) innovative machine learning techniques for dataset evaluation, Hyperparam empowers teams to engineer the highest quality datasets.

Modern LLMs depend on gigabytes of carefully curated, high-quality data. To curate data effectively, engineers need to deeply understand their data. Ideally they should be looking at their data. 👀 But there are no tools that allow this kind of interactive analysis needed for modern ML.

Hyperparam’s mission is to use the browser – the only tool for building modern UIs – to enable engineers to curate massive datasets interactively. By leveraging modern web APIs, and with an obsessive focus on speed and architecture, Hyperparam is building the world’s most scalable UI for dataset curation.

The UI puts the expert in the driver seat, but models give the efficiency gains needed to curate entire datasets single-handedly. We use machine learning models to reflect back on their own training set and assist in finding “good” vs “bad” data. Everyone evaluates models – we evaluate data.

Combine this new scalable UI with methods for evaluating ML data, and you have a powerful engine for iteratively developing the world’s best quality models.