Your models, in the same graphas the services that serve them.
Pull models and experiments from MLflow into UIGraph, then follow a deployed version to its service, infrastructure, cost, and owner without rebuilding the story by hand.
Registered model
checkout-ranking
A registry knows the model. Production needs the whole story.
- 01Which version is in production?
- 02Which service is serving it?
- 03Who owns the path around it?
- 04What did the last evaluation find?
MLflow stays.
Experiments and runs remain in the tracker your ML team already uses.
# Read from the source
ml:
mlflow:
trackingUri: ${MLFLOW_TRACKING_URI}
experiments:
- checkout-ranking
- fraud-scoringOne command refreshes the graphUIGraph joins what MLflow cannot know.
- The service serving it
- The team that owns it
- The infra and cost behind it
The model record
Context accumulates. It never scatters.
From the first imported run to the latest finding, each fact stays attached to one model record.
Experiments and runs
Synced from MLflow while MLflow remains the source of truth.
Cards and versions
Purpose, license, limitations, and every registered version.
What is live
The active version, its environment, and the service exposing it.
Findings stay attached
Evaluation results remain on the model they concern, not in a detached ticket.
Open one model
Leave with the production answer.
Follow a deployed version through real graph relationships instead of assembling the answer across tools.
checkout-ranking
version v3
ranking-service
production
Recommendations
team
Production infra
cost attributed
Continue through the graph
The model is only one entry point.
Built for ML engineers, platform teams, and engineering leads
Frequently asked questions
Still have a question about ML Studio? The docs cover MLflow connections, model sync, and graph relationships in detail.
Read the docsThe next model
Put it in production without losing everything around it.
Sync MLflow once, then follow models through the same graph as the systems around them.