Eventual

Eventual builds data curation for Physical AI. It turns video and sensor data into signals: actions over time, contact and hand–object interactions, trajectories and a quality measure of whether an example is reliable enough to learn from. Each signal links back to the exact moment in the source and comes with a measure of its accuracy.

Customers are frontier Physical AI labs choosing pretraining data and robotics teams improving their models with fleet data. Researchers use the signals to search and inspect behaviours, to select and filter training data, and as supervision targets. Signals are versioned with each data release. The company's open-source engine, Daft, is used for processing, which the website says handles 2 PB of data a day at Amazon.

The team is described as coming from AWS, Lyft and Tesla. The company has raised $30M from Felicis, CRV and Y Combinator, plus angels including the co-founders of Databricks and Perplexity. Open roles are in engineering in San Francisco, including multimodal research, multimodal backend systems, data systems and multimodal data.

Product: Video and sensor data signals (actions, contact, trajectories, quality) for search, filtering and training, plus the open-source Daft data engine.