Voxel51 3.0: Multimodal Data Platform for Physical AI

Voxel51 3.0 is an end-to-end, extensible platform for physical AI teams that unifies multimodal data search, curation, annotation, and model evaluation as systems move from prototypes into production. The release adds support for time-series data across sensor streams, segment-level embeddings for locating relevant moments in long recordings, hybrid semantic, lexical, and metadata search, and unified 2D, 3D, video, and multimodal annotation with quality controls. Its Agent feature combines more than 45 MCP tools, 80 operators, and 18 open-source skills to automate repeatable workflows across the data loop, including finding model failures, curating training examples, generating labels through vision-language models, routing uncertain cases for human review, and evaluating new models. Voxel51 also introduces provenance capabilities such as activity logs, sample edit histories, and immutable dataset versions to support auditing, reproducibility, and governance. Designed to be model-agnostic, the platform allows teams to use proprietary or foundation models for embeddings, labeling, inference, and evaluation, while integrations and custom plugins accommodate existing infrastructure. Voxel51 is offering free access through its hosted application, self-hosted options for teams with data restrictions, and workshops with cloud provider Nebius.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.