Eventual

Eventual is the company behind Daft, an open source multimodal data engine built for AI workloads. Daft processes video, images, audio and sensor data alongside structured metadata in a single dataframe, taking raw data through to training ready datasets in one pipeline. Daft exposes a Python dataframe API and SQL, so anyone familiar with Pandas or Spark can load a dataset, filter, transform, aggregate and write it out. It runs CPU and GPU steps in the same pipeline, scheduling and batching GPU inference or embedding work alongside CPU decoding and filtering. Built in functions let a pipeline call language models and embedding models from providers such as OpenAI, then write results to vector stores like Turbopuffer. Documented use cases include AI search, data enrichment and multimodal ETL. Users bring their own models, storage and infrastructure, and the same code scales from a laptop to distributed clusters at petabyte scale. The audience is data and machine learning engineers, with Eventual specifically offering to help robotics teams curate better data for training. The engine is licensed under Apache 2.0 and installs with pip install daft. Daft is free to use. Eventual does not publish prices for its commercial offering and directs interested teams to request a demo.