Nyckel

Nyckel is a platform that lets developers build custom machine learning functions—such as classification, object detection, optical character recognition, and search—and expose them through a stable API. Instead of managing models or infrastructure, users provide labeled examples, and Nyckel automatically trains, evaluates, and hosts the best model for each task. The platform is designed to solve the problem of making consistent, fast, and accurate AI decisions in production applications, where generative AI is less suited for repeated, deterministic predictions.

Key capabilities include ready-made classifiers and detectors that can be tested immediately, as well as custom models trained on user-provided images, text, or structured data. The platform supports active learning, where models improve automatically from captured production feedback and user review. Nyckel integrates with common development workflows via a REST API and offers features like prediction review, confidence scoring, and private inference. It also provides multi-model decision pipelines, human review queues, and prediction auditing for higher-volume or regulated use cases.

Nyckel is aimed at developers and teams who need to embed AI decisions into their applications without deep machine learning expertise. It offers a free plan with instant access to pretrained models and limited usage. Paid plans are subscription-based and include increasing numbers of invokes (inference requests), samples (stored predictions for training), and functions (individual AI endpoints). All plans are usage-based beyond included limits, with per-invoke and per-sample overage rates. Enterprise plans add dedicated infrastructure, regional deployments, and custom terms.