Kili is a data annotation, AI evaluation, and post-training platform that lets domain experts label, review, and audit training datasets across text, image, video, audio, and geospatial data.
About
Kili is a platform for building, annotating, and evaluating the datasets that machine learning models depend on. It addresses a specific problem in AI development: automated benchmarks cannot tell whether a model is correct in ways that only a trained specialist — a clinician, an imagery analyst, or a domain expert — can judge. Kili brings those experts into structured review and labeling workflows, measures how well they agree, and records every judgment so teams can audit how their data was produced.
The product covers three main use cases: AI evaluation, where specialists judge model outputs; post-training, where experts write the examples used to fine-tune models; and data labeling, where domain specialists annotate assets that only they can classify correctly. Kili is developed by Kili Technology, a company that also publishes documentation, a Python SDK, and an API for programmatic access. The platform supports multiple data modalities and offers cloud, hybrid, and on-premise deployment options for teams with strict data residency requirements.
Key Features
Pricing & Fees
- Pricing: A free trial is available at $0 per month with 2–5 team seats, a 100-asset limit for text, documents, and images, and a 5-asset limit for video and satellite imagery. The Grow plan is a custom subscription with up to 20 seats and up to 50,000 assets, and the Enterprise plan is a custom contract adding SSO, custom terms, and advanced security.
- No credit card required: The free trial and a free two-week evaluation for teams require no credit card.
- Add-ons: Options include expanded cloud storage, additional assets and seats, dedicated labeling services, on-premise deployment (Enterprise-only), and data-regulation compliance support (Enterprise-only).
- API and Python SDK: Available on the Grow plan and above for programmatic integration.
Who It's For
Kili is aimed at machine learning and data science teams that need expert-verified training or evaluation data, particularly in regulated or specialized domains. Typical users include ML engineers fine-tuning models, annotation team leads managing large workforces, AI governance and compliance teams that need auditable records under rules such as the EU AI Act, and specialists in fields like banking, insurance, healthcare, and geospatial analysis. Case studies on the vendor's site include LCL Bank automating KYC document extraction, insurer Covea analyzing 1.5 million customer comments, and Enabled Intelligence achieving over 95% accuracy on millions of geospatial labels.
FAQ
What is Kili?Kili is a data annotation, AI evaluation, and post-training platform that lets domain experts label data, judge model outputs, and produce fine-tuning datasets with a full audit trail.
Is there a free plan?Yes, Kili offers a free trial at $0 per month with 2–5 seats and asset limits of 100 for text, documents, and images and 5 for video and satellite imagery, with no credit card required.
How is Kili priced?Beyond the free trial, the Grow plan is a custom subscription supporting up to 20 seats and 50,000 assets, and the Enterprise plan is a custom contract with SSO, custom terms, and advanced security features.
Can Kili run on-premise?Yes, Kili can run on the vendor's cloud, in a hybrid setup where files stay in the customer's own storage, or fully on the customer's Kubernetes infrastructure; on-premise deployment is an Enterprise-only add-on.
Which data types does Kili support?Kili supports image, video, audio, natural language and LLM text, OCR and document layout analysis, and geospatial imagery annotation.
Related Tutorials
View all →
Related Tools
View all →
Roboflow Roboflow is an end-to-end computer vision platform that lets teams annotate images, train models, and deploy vision AI in the cloud or at the edge. 
PentAGI - Advanced AI-Powered Penetration Testing Fully autonomous AI Agent for complicated penetration testing tasks. 
AI OCR Extract text from images and convert to editable Word documents with AI-powered OCR. 
Keygraph - Continuous Agentic Pentesting for the Enterprise Continuous agentic pentesting and AppSec for the enterprise. Autonomous exploits validated against your live app, inside your perimeter. Built by the Shannon team. 
Evidently AI - AI Evaluation & LLM Observability Platform Evidently AI is a platform for AI evaluation and LLM observability, helping teams monitor and improve model performance.