Automating Complex and Distributed Forecasting Tasks for Faster Supply Chain Decisions

Abstract

Forecasting is a key AI component that drives various supply chain use cases such as inventory management, markdown optimization, etc. In general, supply chain use cases deal with large-scale data that needs sophisticated distributed forecasting techniques. These techniques involve a lot of complex steps such as pipeline construction, set-up/execution across multiple distributed environments (ray, spark), HPO, right model selection, backtesting, evaluation, etc. Manually coding and orchestrating these tasks is highly time-consuming and error-prone. To tackle this, we propose our in-house built YAML-driven orchestration engine that automates and eases various complex distributed forecasting tasks for faster supply chain decisions.