Encord

Last updated: 2026-08-12

Encord is a multimodal data platform used by 300+ AI teams, including Woven by Toyota, AXA, UiPath, and Skydio, to annotate, curate, and evaluate training data. Founded in 2021 and backed by $110M in funding, it supports images, video, audio, text, DICOM medical imaging, and LiDAR/3D point clouds with SOC 2, HIPAA, and GDPR compliance.

About Encord

Encord is a data development platform for AI teams, built by Ulrik Stig Hansen and Eric Landau and backed by $110M in total funding, with Wellington Management among its investors. It gives ML engineering and data teams one place to curate, annotate, and evaluate the training data behind computer vision and multimodal models, positioning itself against cloud-first labeling vendors as infrastructure built for physical AI: robotics, autonomous vehicles, and smart infrastructure rather than just web and app data. The platform is organized around three connected modules. Encord Annotate handles labeling spanning pictures and footage, spoken audio and written text, scanned documents, medical imaging in DICOM/NIfTI format, and 3D LiDAR point clouds, with bounding box, polygon, polyline, keypoint, bitmask, and semantic segmentation tools. Encord Active adds curation and active learning so teams can catch bad labels and weak training examples early in a project, and Index handles search and discovery across large multimodal datasets without exporting them elsewhere. AI-assisted labeling built on SAM 2 and GPT-4o generates first-pass labels for images and video, and a Python/REST SDK lets teams script datasets, ontologies, and pipelines into existing MLOps stacks rather than working only through the web UI. Encord is built for ML engineers and data operations teams on computer vision and physical AI programs: robotics and autonomous vehicle labeling, healthcare AI teams working with DICOM imaging, and enterprise teams standardizing training-data pipelines across multiple AI projects. Customers named on Encord's site include Woven by Toyota, AXA, UiPath, Zipline, Skydio, and Maxar, and it is SOC 2 and HIPAA certified and meets GDPR requirements, with VPC and on-premises deployment for teams that cannot use a public SaaS labeling tool. Every plan requires a sales demo rather than a published price, a contrast with self-serve competitors like Roboflow and SuperAnnotate. Encord previously offered an open-source curation toolkit, Encord Active, on GitHub, but that standalone repository was archived in August 2025 and its functionality now lives inside the paid Active Cloud product.

Pricing

Enterprise-only pricing across three tiers, Starter, Team, and Enterprise; none publish a list price and all require a sales demo. Starter targets individuals and small teams prototyping AI applications, Team adds active learning pipelines and performance analytics for teams scaling a few AI applications, and Enterprise adds SSO, dedicated support, and custom SLAs for organizations running multiple AI projects.

Key Features

• Multimodal annotation: Labels seven data types (images, video, audio, text, documents, DICOM/NIfTI medical imaging, and LiDAR/3D point clouds) in one platform using bounding boxes, polygons, polylines, keypoints, bitmasks, and semantic segmentation.
• AI-assisted labeling: SAM 2 and GPT-4o integrations auto-generate first-pass labels on image and video data, cutting manual frame-by-frame annotation work.
• Encord Active for curation: Active learning pipelines and error discovery flag mislabeled or low-value samples before they reach a model, closing the loop between annotation and evaluation in one workspace.
• Index for large-scale data discovery: Searches and explores multimodal datasets at scale without exporting them to a separate tool, paired with Data Agents that automate repetitive pipeline steps.
• Customizable SDK: A Python/REST SDK scripts datasets, ontologies, projects, and labels, letting teams wire Encord into existing MLOps pipelines instead of working only through the web UI.

Pros

• Handles seven data modalities, including DICOM/NIfTI and LiDAR/3D, in one platform, avoiding the multi-vendor stitching common with single-modality labeling tools.
• SOC 2 Type II, HIPAA, and GDPR compliance plus VPC and on-premises deployment options make it usable for healthcare and regulated teams that cannot send data to a public SaaS labeling tool.
• Rated 4.8/5 across 65 reviews on G2, with reviewers citing support tickets resolved within 1-2 days and ease of managing large annotation teams.
• Raised $110M total, including a $60M Series C led by Wellington Management in February 2026 at a $550M valuation, giving it more runway than most single-modality annotation vendors.

Cons

• No public pricing: all three tiers (Starter, Team, Enterprise) require a sales demo, unlike per-seat competitors such as Roboflow or SuperAnnotate that publish self-serve pricing.
• Encord Active, its formerly open-source curation toolkit, was archived on GitHub in August 2025, folding that functionality into the paid Active Cloud product and removing the free self-hosted option.
• A 150-person team (per its YC profile) supporting seven data modalities across a growing enterprise customer base can mean longer onboarding queues for large datasets.

Visit Encord Official Website