Workshops Use Cases Architecture Projects Blog
About Contact Offers Talks
Home›AI technologies›Your own modelDeutsch

Your own model: train it, version it, serve it

When no ready-made service and no prompt does the job, you train or fine-tune a model on your own data and serve it yourself. An ML platform gives you the same six steps everywhere.

Amazon SageMaker AIAzure Machine LearningVertex AIRed Hat OpenShift AI, Kubeflow and MLflow
Book a free 30-min call →Workshop: OpenShift AI on AWS lab →

What does running your own model involve?

The six steps are always the same: develop in a notebook, train or fine-tune on your data, register the result as a version, deploy it to an endpoint or a batch job, monitor it, and evaluate whether it is still good enough on this month's data. A model is never shipped once.

Training is temporary compute that ends by itself. A real-time endpoint is persistent compute that bills while it waits. Owning the weights does not have to mean owning a machine: a small tuned model can also be imported into a managed service and billed per use.

Goes inData, code, a training job, later a model to deploy
Comes outA trained model, a registry entry and an endpoint
Typical usesKnowledge no public model has, strict data rules, specialised predictions

Four ways to run it

Each of the big clouds offers it as a managed service, and you can also run it yourself. Same task, a different bill and a different console.

Amazon SageMaker AI
Managed on AWS

Amazon SageMaker AI[1]

The managed machine learning platform on AWS: notebooks, training jobs, a model registry and endpoints.

Official documentation →
Azure Machine Learning
Managed on Microsoft Azure

Azure Machine Learning[2]

The managed machine learning platform on Microsoft Azure.

Official documentation →
Vertex AI
Managed on Google Cloud

Vertex AI[3]

The managed AI platform on Google Cloud, with training, a model registry and endpoints.

Official documentation →
Red Hat OpenShift AI, Kubeflow and MLflow
Run it yourself

Red Hat OpenShift AI, Kubeflow and MLflow[4]

The platform you run yourself, on premises or on a cloud. OpenShift AI bundles open-source components such as PyTorch, Kubeflow, MLflow and vLLM, with Red Hat support; Kubeflow and MLflow can also be assembled without it.

Project page →
Amazon SageMaker AI: The managed service on one page, with the same service on the other platforms.
The managed service on one page, with the same service on the other platforms.[5]

How it looks in an architecture

Slides from my course on AI architectures at FH Burgenland, drawn from real projects.

Your own data and your own model on your own endpoint: explore, train, version, serve.
Your own data and your own model on your own endpoint: explore, train, version, serve.[6]
Where a model of my own lives when there is no server of mine: trained on a rented GPU, stored in S3, served by Bedrock Custom Model Import.
Where a model of my own lives when there is no server of mine: trained on a rented GPU, stored in S3, served by Bedrock Custom Model Import.[5]
OpenShift AI: the same operator on whichever OpenShift you have, managed or self-managed.
OpenShift AI: the same operator on whichever OpenShift you have, managed or self-managed.[6]
Red Hat OpenShift AI on one page: what you send, what you get back, and who operates what.
Red Hat OpenShift AI on one page: what you send, what you get back, and who operates what.[5]

Where I used it

The event assistant

My event assistant needed knowledge no public model has: the group's own history. I fine-tuned a small open model with InstructLab on a rented GPU for a few hours, stored the weights in S3 and serve them through Bedrock Custom Model Import. A SageMaker endpoint was considered and rejected, because it bills for every hour it exists. The whole system runs at about thirty euro a month.

See the reference projects →

Managed and self-hosted together: hybrid

With your own model, the model artefact is yours, and the training data may not be allowed to leave. That splits the work: training where the data is, serving where the users are. On your own platform, hybrid means the same platform installed twice, on machines you own and machines you rent, run as one. Red Hat supports OpenShift AI Self-Managed on the managed OpenShift offerings as well as on your own hardware.[7]

Two grounds under one operating model: a managed model in a region, an open model on OpenShift AI in your building.
Two grounds under one operating model: a managed model in a region, an open model on OpenShift AI in your building.[6]

What would help you next?

Learn to decide

The course “From Managed to Hybrid” teaches where AI should run: what each layer costs, what you operate and where your data has to stay. It is in preparation, with a free live session first.

Join the course waitlist →
Linda MohamedLinda Mohamed
AI and cloud consultant, AWS Community Hero, Vienna

Build it with me

Tell me what you want to build. In 30 minutes we check which of the four ways fits your data, your rules and your budget, and whether AWS funding for a proof of concept can cover part of it. How AWS funding works →

Book a free 30-min call →15-min question slot

You already know what you need

Book the workshop directly:

ai-solutions.wiki

Read more on ai-solutions.wiki

My open knowledge base, with longer comparisons, pricing notes and sources.

Sources

  1. Amazon SageMaker AI documentation. https://docs.aws.amazon.com/sagemaker/latest/dg/whatis.html
  2. Azure Machine Learning documentation. https://learn.microsoft.com/en-us/azure/machine-learning/overview-what-is-azure-machine-learning
  3. Vertex AI documentation. https://cloud.google.com/vertex-ai/docs
  4. Red Hat OpenShift AI, Kubeflow and MLflow. https://www.redhat.com/en/products/ai/openshift-ai
  5. Linda Mohamed, course slides “The projects”, MASE-3, FH Burgenland, September 2026
  6. Linda Mohamed, course slides “How to run AI, layer by layer”, MASE-3, FH Burgenland, September 2026
  7. Red Hat OpenShift AI supported configurations 3.x, Red Hat. https://access.redhat.com/articles/rhoai-supported-configs-3.x
  8. Amazon SageMaker, ai-solutions.wiki. https://ai-solutions.wiki/tools/amazon-sagemaker/
  9. Red Hat OpenShift AI, ai-solutions.wiki. https://ai-solutions.wiki/tools/openshift-ai/
  10. SageMaker vs Bedrock, ai-solutions.wiki. https://ai-solutions.wiki/comparisons/sagemaker-vs-bedrock/
  11. Custom ML models vs foundation models, ai-solutions.wiki. https://ai-solutions.wiki/comparisons/custom-ml-vs-foundation-models/
  12. Bedrock Custom Model Import, AWS documentation. https://docs.aws.amazon.com/bedrock/latest/userguide/model-customization-import-model.html
  13. InstructLab. https://instructlab.ai/

Links checked September 2026. Product names and features change: check the linked pages before you decide.