Understand and prepare your data
Check data quality, prepare useful inputs and keep the dataset version behind every model.
From data to deployed AI. One platform.
Train, evaluate, deploy and monitor machine-learning models in one platform. Give your applications a prediction API without assembling an entire ML platform yourself.
- Self-hosted
- Your infrastructure
- Your models
Data → Train → Evaluate → Deploy → MonitorYour team stays in control
01Add your dataCheck quality and prepare inputs
02Train modelsAutomatically train and compare
03EvaluateExplain results and approve a model
04DeployGive applications a prediction API
05MonitorTrack health and investigate changes
The outcomeA model your applications can use. A lifecycle your team can manage.
You approve the modelSceptre connects the steps
For data and analytics teams that want to use machine learning without building a platform first.
Why Sceptre existsEverything between the notebook and production.
Profile data, prepare features, train and tune models, compare results and operate the model you choose. Sceptre reduces repetitive platform work so data scientists can focus on the decisions that need their expertise.
Train and compare models
Train multiple approaches automatically. Tune settings, compare experiments and evaluate the results that matter to your task.
Explain and approve the choice
Test on separate data, understand what influenced predictions and give stakeholders evidence they can review.
Deploy and monitor your model
Record the model you approve, deploy it as a prediction API and track its health as new data arrives.
Machine learning shouldn’t require six different platforms.
Platform approach
Build your own ML platform
What your team takes on
Tailor every part to your needs. Your team builds and maintains the integrations, deployment services, monitoring and upgrade paths.
From business data to usable predictions.
A model learns patterns in your existing data, then uses them to predict outcomes for new data. Sceptre connects training, review, deployment and monitoring in one workflow.
Read the quick start01
Choose with confidence
Add your data, check its quality and train models automatically. Compare results, test on separate data and review explanations before approving a model.
02
Turn your model into an API
Business application → Sceptre prediction API → Model → Prediction. Your systems send data and receive a prediction, such as churn risk, a lead score or a demand estimate.
03
Deployment isn’t the end
Track the deployed version and service health. Investigate changes in incoming data, called data drift. Check accuracy against actual outcomes when available, then decide whether to retrain or replace the model.
04
Your data. Your infrastructure. Your models.
Run Sceptre on infrastructure your organisation controls. Keep sensitive data and model operations in your environment, reduce platform dependency and align deployment with your governance requirements.
Know why the model was chosen
Follow each model from its data to deployment and monitoring.
Know how every model reached production.
Give technical teams, business leaders and risk reviewers a shared record of the data, experiments, evaluations and approvals behind each deployed model.
- Understand which inputs influenced predictions and review the evidence behind the choice.
- Trace the data versions, tests and approval attached to each model.
- See which models were deployed, monitored and replaced, with your team accountable for each decision.
Your team focuses on models. Sceptre connects the lifecycle.
Give data and analytics leaders one view of how projects move from experiments to prediction services. Help smaller teams deliver without assembling a large ML platform team.
Move faster
Take a dataset to a usable prediction service without building every part of the ML platform yourself.
Reduce platform complexity
Bring disconnected tools into one workflow, with fewer integrations and systems to maintain.
Maintain control
Keep data, infrastructure and models under your organisation’s control. Your team owns the deployment decisions.
Know what is running
See which model was approved and deployed, why it was chosen and when it needs attention.
Know exactly where Sceptre fits.
Practical answers about ownership, architecture and getting started.
What is Sceptre?
Sceptre is an end-to-end machine-learning platform for business data in rows and columns. Train, compare, evaluate and approve models, deploy them as prediction APIs, and monitor them in one place.
Who is Sceptre for?
Data and analytics leaders, data scientists, ML engineers and governance teams that want one workflow from data to deployed models. Smaller businesses can avoid assembling an entire ML platform, but still need the skills or support to operate their infrastructure.
Is Sceptre self-hosted?
Yes. Your data, infrastructure and models remain under your organisation’s control. Sceptre is Kubernetes-native: your cluster runs training jobs and prediction services. Your team manages access, capacity and maintenance as workloads grow.
What is underneath the platform?
Sceptre combines automated model training (AutoML) with model deployment and operations (MLOps). Versioned data and model files use S3-compatible storage. MLflow tracks experiments, a registry records model versions, and Kubernetes runs jobs and authenticated prediction services. Install with the public Helm chart and review the technical documentation for requirements.
How does Sceptre explain model decisions?
See which inputs influenced a prediction and give reviewers evidence behind the model choice. The technical method is called SHAP. Explanations support review; they do not prove that a model is correct or fair.
What can my business use predictions for?
Explore customer churn, lead scoring, customer propensity, demand, pricing, claims or fraud and credit risk. Your application calls the deployed API with new data and uses the prediction in its workflow. Each use case needs suitable data, validation and human oversight.
What data and model types can I use?
Bring CSV, Parquet, Excel, JSON or JSONL files within the documented size limits. Predict categories (classification), estimate numbers (regression), group similar records (clustering) or model observations over time (time series). Sceptre focuses on tabular machine learning, not general-purpose LLM training.
How do I know when a model needs attention?
Review service health, the version deployed and changes in incoming data. Data drift is a signal to investigate. Measuring changes in prediction accuracy also requires actual outcomes for comparison. Your team reviews the evidence and decides whether to retrain or replace the model.
Is Sceptre ready for production use?
The current release is for local development, evaluation and compatibility testing. Shared or production use requires qualification against the production-readiness guide. Start Building takes you to the self-hosted evaluation instructions.
Your next model shouldn’t stop at the notebook.
Train it. Evaluate it. Explain it. Approve it. Deploy it. Monitor it. From data to production, all from Sceptre.