Foresight Inference API
Scalable machine learning inference API for time-series forecasting and demand prediction.
Overview
A fully managed forecasting engine built for time-series data. Foresight API automatically selects, trains, and evaluates multiple statistical and deep learning models (ARIMA, Prophet, DeepAR, Transformer-based) to provide the most accurate predictions for your specific dataset.
Key Features
AutoML model selection
Automatically identifies the best algorithmic approach per series.
Covariate support
Include external factors (weather, holidays, promotions) to improve accuracy.
Probabilistic forecasting
Returns confidence intervals (p10, p50, p90) rather than point estimates.
Common Use Cases
- Retail inventory optimization
- Energy grid load forecasting
- Dynamic pricing adjustments
Technical Specifications
Architecture
Serverless inference endpoints
Scalability
Autoscales to 0, bursts to 10k concurrent requests
Data Ingestion
S3, GCS, direct payload
Latency
Optimized for batch and real-time processing