Build AI you can trust and truly own — datasets and models crafted for clarity, performance, and independence.
Every resource is designed to help small teams deploy private, reliable AI faster, without relying on third-party APIs or black-box tools.
Each dataset is clean, validated, and stratified, ensuring balanced performance and reproducible metrics (macro-F1, recall, accuracy). Our models are trained on the very same data, ready to deploy through Hugging Face or ONNX.
No calls to third-party APIs, no telemetry, no data leaks. You download, inspect, and deploy — your information never leaves your infrastructure.
Most open datasets target English.
Kinoux fills the gap with French-native corpora — reviews, emails, forms, dialogues — and models adapted to local syntax and tone.
All assets ship with schema notes, label dictionaries, quick-start scripts, and example inference code. Plug the model into your forms, inbox, or app in hours — not weeks.
Get line-delimited JSONL corpora with labels, splits, and evaluation templates. Ideal for ML engineers and data scientists who want control, auditability, and reusability.
Clean data saves you hundreds of hours of collection, cleaning, and annotation.
Deploy fine-tuned, ready-to-use Transformers trained on our datasets. Ideal for developers who want plug-and-play results without GPU or ML setup.
Skip the training stage — integrate, test, and deliver value today.
Gain production-ready AI without recruiting data scientists.
Access clean, domain-balanced datasets.
Embed reliable NLP without cloud lock-in.
Below you’ll find our releases — each combining a dataset and its matching fine-tuned model ready for local use.
