Learn
In-depth guides, core concepts, and best practices for getting the most out of Nyckel. The Get Started tab gets you to a working function; this tab takes you a level deeper into the ideas behind the platform and the practices that make a function actually good over time.
Core Concepts
The handful of ideas the rest of the product is built on.
Functions
The core unit: model + samples + review + pipeline.
Models
What's inside a function and how Nyckel selects and updates them.
Labels
The decision space of a classification function.
Samples
The labeled examples that train private models.
Confidence Scores
What the score means and how to read it.
Invoke Capture
How informative production invokes reach the Review queue.
Best Practices
How to design labels, run a review workflow, and tune confidence thresholds — the practical decisions that determine how good your function gets.
Get Reliable Predictions
Designing labels and data for trustworthy predictions from day one.
Review and Improve Predictions
Build a sustainable review workflow that compounds over time.
Improve Accuracy Over Time
Feedback, annotations, and targeted samples.
Tuning Confidence Thresholds
Picking the cutoffs your application acts on.