Curriculum / Data Scientist
Data Scientist on Foundry
Model and forecast supply-chain disruption risk: build a feature dataset, train a model, evaluate it under a Modeling Objective, analyze and visualize results (Contour + Quiver), score the Ontology objects via batch inference, stand up a live deployment for on-demand inference, bind model outputs to the Ontology via a published model Function, and operationalize the forecast end-to-end.
Data scientists (coding-capable — Python modeling + Code Repositories/Workspaces).
The 8-unit arc
- Unit 1: Frame the modeling problem & build the feature datasetintermediate
Define the disruption-risk prediction target and build a feature-engineered, labeled training dataset with a held-out test split.
- Unit 2: Train a disruption-risk model in Code Repositories / Code Workspacesintermediate
Train and publish a Foundry model asset (artifact + adapter) with version history and training-dataset lineage.
- Unit 3: Evaluate the model under a Modeling Objectiveadvanced
Create a Modeling Objective containing the submitted model and a MetricSet over the held-out evaluation dataset.
- Unit 4: Statistical analysis & visualization of model resultsadvanced
Quantify error/segment risk in Contour and visualize predicted disruption risk over time in a Quiver dashboard.
- Unit 5: Batch inference: score the Ontology objectsadvanced
Produce a batch-predictions dataset and bind a DisruptionRisk property onto Supplier objects in the Ontology.
- Unit 6: Live deployment for on-demand inferenceadvanced
Stand up a running live deployment exposing the model behind a queryable REST inference endpoint.
- Unit 7: Bind model outputs to the Ontology via a published model Functionadvanced
Publish a model Function (query function) taking object-property inputs and returning a disruption-risk prediction, callable from Actions/Workshop.
- Unit 8: Capstone: operationalized disruption-risk forecast in the Ontologycapstone
An end-to-end governed capability: Ontology objects carry a model-derived DisruptionRisk, a published model Function powers on-demand scoring, with documented lineage and monitoring.
Frame the modeling problem & build the feature dataset
Define the disruption-risk prediction target and build a feature-engineered, labeled training dataset with a held-out test split.
Train a disruption-risk model in Code Repositories / Code Workspaces
Train and publish a Foundry model asset (artifact + adapter) with version history and training-dataset lineage.
Evaluate the model under a Modeling Objective
Create a Modeling Objective containing the submitted model and a MetricSet over the held-out evaluation dataset.
Statistical analysis & visualization of model results
Quantify error/segment risk in Contour and visualize predicted disruption risk over time in a Quiver dashboard.
Batch inference: score the Ontology objects
Produce a batch-predictions dataset and bind a DisruptionRisk property onto Supplier objects in the Ontology.
Live deployment for on-demand inference
Stand up a running live deployment exposing the model behind a queryable REST inference endpoint.
Bind model outputs to the Ontology via a published model Function
Publish a model Function (query function) taking object-property inputs and returning a disruption-risk prediction, callable from Actions/Workshop.
Capstone: operationalized disruption-risk forecast in the Ontology
An end-to-end governed capability: Ontology objects carry a model-derived DisruptionRisk, a published model Function powers on-demand scoring, with documented lineage and monitoring.