Kili Technology

Provides a visual, drag-and-drop interface for labeling data without requiring coding knowledge. Supports multiple annotation types including classification, bounding boxes, polygons, and semantic segmentation.

Automatically calculates agreement metrics between multiple annotators on the same data samples. Identifies labeling inconsistencies and flags low-confidence annotations for review.

Provides programmatic access to annotation data, project management, and quality metrics through REST endpoints. Enables custom integrations with ML pipelines and external tools.

Manages user accounts, roles, and granular permissions across projects and datasets. Supports SSO integration and role-based access control (RBAC).

Generates dashboards and reports on annotation progress, quality metrics, annotator performance, and project timelines. Includes visualizations and exportable reports.

Provides features for data anonymization, encryption, access logging, and compliance with regulations like GDPR and HIPAA. Includes data retention policies and secure deletion.

Calculates statistical measures (Cohen's kappa, Fleiss' kappa, Krippendorff's alpha) to quantify agreement between annotators. Provides detailed breakdowns by sample, category, and annotator.

Enables teams to assign annotation tasks, track progress, set deadlines, and manage workload distribution across multiple annotators. Includes role-based access control and task routing.

Allows creation of reusable annotation schemas and templates for common labeling tasks. Supports conditional logic, custom fields, and predefined label hierarchies.

Handles bulk uploading of datasets in multiple formats and organizes them into projects. Supports incremental imports, deduplication, and metadata association.

Exports labeled data in multiple formats compatible with ML frameworks (COCO, Pascal VOC, YOLO, etc.). Supports custom export schemas and filtering.

Tracks changes to annotations over time, maintains version history, and allows rollback to previous annotation states. Includes audit logs of who changed what and when.

Identifies and prioritizes the most informative or uncertain samples for annotation, reducing labeling effort. Uses model predictions or uncertainty metrics to rank samples.

Implements multi-stage review processes where senior annotators or QA reviewers can approve, reject, or request changes to annotations. Includes feedback mechanisms and revision tracking.