AI Architecture › MLOps & Learning
MLOps & Learning
MLOps & Learning
Offline retraining, twin refinement, drift detection, and reinforcement learning from human feedback.
What this layer does
Offline retraining, twin refinement, drift detection, and reinforcement learning from human feedback. Every page in this section is part of the same control stack — click any card to go deeper, or return to the Master Blueprint to see how it connects.
Explore this layer
A/B and Interleaving Evaluation OfflineActive Learning and Experiment DesignCI/CD for Model PipelinesCanary Rollout of Control PoliciesChampion-Challenger EvaluationConcept Drift DetectionData Versioning and LineageDetecting Model Degradation in ProductionDiagnostic Calibration Drift and RecalibrationDigital-Twin RefinementDrift Detection: Covariate ShiftEdge Model Compilation and DeploymentExperiment Tracking and MetadataFeature Store for Plasma and Diagnostic SignalsFleet Model Propagation Across UnitsGovernance and Approval WorkflowHonest Gates in the MLOps PipelineHyperparameter Optimization and Architecture SearchIncident-Driven Retraining and PostmortemsLabel Quality and ProvenanceMLOps Lifecycle for Fusion ControlModel Cards and Dataset DatasheetsModel Registry and Promotion StatesModel Versioning and LineageOffline Batch RL from Logged PulsesOffline Retraining on L0Preference Data Collection and CurationPreference-Based RL for Control PoliciesPreventing Training-Serving SkewProduction Model MonitoringQuantization and Pruning for the EdgeRLHF for Operator CopilotsRegime Drift Detection on Plasma StateReproducible Training RunbooksReward Modeling for Control and CopilotsRollback and Safe-State FallbackShadow Deployment of ControllersSim-to-Real Transfer and Domain RandomizationSurrogate Retraining CadenceThe Continual Learning LoopValidation Gates Before a Model Touches the Machine