Nyckel consultants

Q: How much training data does Nyckel need to produce a useful model?

Nyckel can start training with just a few labelled examples per category. Accuracy improves as you add more data, but you can test a working model quickly and iterate. For most production use cases, a few dozen to a few hundred labelled samples per category produces solid results. The platform tells you when more data would improve accuracy.

Q: How does Nyckel compare to building a custom ML model from scratch?

Building from scratch gives you maximum flexibility but requires data science skills, GPU infrastructure, and weeks of development time. Nyckel trades some flexibility for speed: you get a production-ready classification API in minutes rather than months. For standard classification tasks (sorting, categorising, moderating), Nyckel covers the use case without the overhead.

Q: Can we integrate Nyckel's classification API into our existing application?

Yes. Nyckel provides a REST API that you call from any application that can make HTTP requests. Our integration team connects Nyckel to client systems using direct API calls, automation platforms like n8n, or serverless functions. This means classification can run inline within your app or as a batch process on queued items.

Q: Does Nyckel's model improve automatically over time?

Nyckel retrains your model automatically when you add new labelled data. This means you can correct misclassifications and add new examples through the web interface or API, and the model improves without you needing to manage retraining infrastructure. It is a continuous improvement cycle: use the model, review edge cases, add corrected labels, model gets better.

Q: What does Nyckel cost and how is pricing structured?

Nyckel uses pay-per-invocation pricing, meaning you pay per API call rather than for model training or hosting upfront. This makes it cost-effective for variable workloads: you are not paying for idle GPU time during quiet periods. Check Nyckel's current pricing page for specific rates, as they may vary by classification type and volume.

We can help you automate your business with Nyckel and hundreds of other systems to improve efficiency and productivity. Get in touch if you’d like to discuss implementing Nyckel.

Get in touch

Book a call

About Nyckel

Nyckel is a machine learning API that lets developers build and deploy custom classification models without needing data science expertise. You provide labelled examples (images, text, or tabular data), Nyckel trains a model automatically, and you get an API endpoint you can call from your application within minutes rather than months.

The problem Nyckel solves is the gap between wanting ML classification and actually building it. Training a custom image classifier or text categorisation model traditionally requires a data scientist, significant compute resources, and weeks of development time. Nyckel compresses that process: you upload training samples through a web interface or API, the platform handles model selection, training, and hosting, and you get a production-ready endpoint immediately.

Key features include:

Custom image, text, and tabular data classification models
No-code web interface plus a full REST API for developer integration
Models train in minutes from as few as a handful of labelled examples
Automatic model improvement as you add more training data over time
Hosted inference API with built-in scaling
Semantic search and content moderation functions
Pay-per-invocation pricing with no upfront model training costs

Nyckel is used by development teams that need to add classification to their products quickly: content moderation, document sorting, product categorisation, image tagging, and similar tasks. At Osher Digital, our custom AI development team integrates Nyckel into client applications and automation pipelines, connecting its classification API to business workflows where sorting or categorising data manually creates bottlenecks. We used similar classification approaches in our medical document classification project.

Nyckel FAQs

Frequently Asked Questions

Common questions about how Nyckel consultants can help with integration and implementation

What types of classification can Nyckel handle?

Nyckel supports three main input types: images, text, and tabular data. For images, you can build classifiers that identify objects, detect defects, categorise products, or moderate content. For text, it handles sentiment analysis, topic categorisation, spam detection, and ticket routing. Tabular classification works with structured data like spreadsheet rows for lead scoring or risk assessment.

How much training data does Nyckel need to produce a useful model?

How does Nyckel compare to building a custom ML model from scratch?

Can we integrate Nyckel’s classification API into our existing application?

Does Nyckel’s model improve automatically over time?

What does Nyckel cost and how is pricing structured?

How it works

We work hand-in-hand with you to implement Nyckel

As Nyckel consultants we work with you hand in hand build more efficient and effective operations. Here’s how we will work with you to automate your business and integrate Nyckel with integrate and automate 800+ tools.

Step 1

Process Audit

We review where classification tasks exist in your business: document sorting, content moderation, product tagging, image categorisation, or data triage. We assess current manual processes, estimate classification volumes, and identify which tasks are strong candidates for automated ML classification with Nyckel.

Step 2

Identify Automation Opportunities

From the audit, we prioritise classification use cases by business impact and data readiness. Tasks with clear categories, existing labelled examples, and high manual effort are ideal starting points. We also identify where Nyckel classification can feed into downstream automation, such as routing documents to different teams based on type.

Step 3

Design Workflows

We design the classification pipeline: how data enters Nyckel (API call, batch upload, or automation trigger), what happens to classified items (routed, tagged, flagged), and how misclassifications are corrected to improve the model. We also plan the integration between Nyckel’s API and your existing application or automation platform.

Step 4

Implementation

We set up Nyckel with your initial training data, train the classification model, and integrate the API endpoint into your application or workflow. We configure automated triggers using n8n or direct API calls, set up result handling logic, and establish the feedback loop for correcting misclassifications.

Step 5

Quality Assurance Review

We test classification accuracy using held-out samples from your data, measuring precision and recall across all categories. We identify edge cases where the model struggles, add corrective training data, and verify that the end-to-end pipeline (input, classification, action) works reliably under production conditions.

Step 6

Support and Maintenance

After launch, we monitor classification accuracy weekly, add new training data for misclassified items, and expand the model as new categories emerge. We track API usage and costs, review accuracy metrics monthly, and adjust the pipeline as your classification needs evolve.

Transform your business with Nyckel

Unlock hidden efficiencies, reduce errors, and position your business for scalable growth. Contact us to arrange a no-obligation Nyckel consultation.

Get in touch

Book a call