Encord

Encord is a data infrastructure platform for teams building AI models on multimodal data. It handles the full pipeline from pretraining through deployment: annotating video, images, LiDAR, audio, text and fused sensor streams; curating datasets with embedding search and model in the loop selection of edge cases; evaluating models with rubric scoring, pairwise comparison and RLHF orchestration; and monitoring production models so that failures are routed back into training. The platform is positioned for physical AI teams, meaning robotics, autonomous vehicle, drone and humanoid developers, and for enterprise computer vision in healthcare, defense, smart cities and video intelligence. Capabilities that matter include native video annotation with label lineage and quality controls, complex ontologies, customizable review workflows, data agents that automate labeling and curation steps, distribution gap analysis for finding rare cases, and facility or fleet data collection. Named customers include UiPath, Toyota Research Institute, Agility Robotics and Zipline. Pricing is published as three tiers without amounts. Starter is for individuals and small teams prototyping and includes the image and video annotation toolkit, complex ontologies, customizable workflows and self serve support. Team adds data agents, performance analytics, model evaluation and onboarding support. Enterprise adds multiple workspaces, single sign on, an enterprise SLA and VPC or on premises deployment, and is sold through sales. Starter and Team are self serve sign ups.