Open-source RLHF platform

Ship better AI with human feedback at scale

The annotation platform purpose-built for RLHF. Collect preference data, ensure quality with gold scoring and IAA, and automate with LLM-as-judge — all in one tool.

Multi-format Annotation

Comparison, rating, and ranking tasks with streaming AI response simulation. Purpose-built for RLHF preference data.

Quality Assurance

Gold-standard scoring, calibration tests, inter-annotator agreement, and real-time drift detection.

LLM-as-Judge

Automated batch evaluation with human override. Scale annotation quality without scaling headcount.

Dataset Versioning

Version, diff, and export annotated datasets in JSONL, CSV, or HuggingFace format. Full audit trail.

Team & RBAC

Multi-org support with admin, reviewer, and annotator roles. Bulk assignment and review workflows.

API-first + SDK

Full REST API, Python SDK, and CLI. Automate pipelines with webhooks and API keys.

3
Annotation formats
16+
API endpoints
100%
Open source
SDK
Python + CLI