Q1 2026: Scalability & Operations
Kubernetes Operator
Migrate from Docker Compose to a Helm-managed Kubernetes deployment (EKS).
GPU Support
Enable
task_type="GPU" for CatBoost to reduce optimization time by 50%.Model Registry UI
Deploy MLflow UI behind OAuth proxy for broader team access.
Q2 2026: Advanced Modeling
Deep Learning
Introduce LSTM or Transformer-based model into the A/B testing rotation.
External Regressors
Integrate holiday calendars and major local event data into the feature store.
Real-time Serving
Move from Batch Inference to low-latency Online Inference (gRPC).
Q3 2026: Infrastructure & Resilience
Auto-Scaling Inference
Horizontal Pod Autoscaling (HPA) for the Model Server based on request volume.
Shadow Mode
Deploy “Shadow” models that receive traffic but don’t return predictions (for safe testing).