Rerun

Rerun is a data layer for physical AI, covering robotics, spatial computing and embodied AI. It is built for multi rate, multimodal data such as sensor streams, video and teleoperation logs, and it supports the loop of collecting, refining, training on and evaluating that data. The open source Rerun SDK is the core of the product. It provides logging, columnar storage in the rrd file format, dataframe and SQL queries, a declarative visualization framework, an interactive viewer on desktop and web, and a PyTorch dataloader. It has Python, Rust and C++ APIs and is licensed under Apache 2.0 and MIT. Rerun Hub is the commercial backend: a managed catalog with byte range indexing over a team's own S3 compatible object storage, support for transforms that add derived columns without copying data, streaming of dataset mixes straight to GPUs, link sharing, and single sign on. Hub runs single tenant in a region the customer chooses. It is aimed at robotics and physical AI teams who want to inspect failures, curate training data and train on recordings without building export pipelines. Rerun Technologies AB is a remote first Swedish company that develops in the open. The SDK is free forever. Hub is priced per deployment based on team size and data scale, with contracts arranged through a meeting with the vendor. Academic and research discounts are available on request.