Implemented a ML framework for optimizing delivery routes and efficiencies for fleet utilization, along with inventory planning and management


The client is a Chicago based healthy food delivery start-up with hub-and-spoke distribution model. They have a web/app platform that allows users to order healthy food through pre-orders or on-demand orders.

  1. Inventory planning: Client had problems with inventory planning and management due to lack of foresight into upcoming demands 
  2. Driver allocation: Due to lack of insight into demand and non-scientific methods of driver allocation, there was severe under/over utilization of drivers, leading to significant delivery cost and delay in orders
  3. Long lead time in converting data into insights for the leadership team to understand and measure business performance around metrics like revenue, orders, items, customers, etc.


  1. Built and deployed a demand forecasting model using historical order patterns, for estimating the number of meals (demand) across time slots (lunch, dinner, peak/shoulder demand etc.) and zones

  2. Built and deployed a dynamic driver allocation engine to allocate optimum number of drivers in different zones, across various time slots, based on forecasted demands


Ensured operational effectiveness: Demand forecasting model estimated in advance, upcoming demands with 95% accuracy, helping the client with kitchen and inventory planning

Reduced delivery costs by 30% by improving driver allocation and route optimization

Insights platform helped the leadership team track performance of KPIs like no. of customers, items, orders, revenue and recent demand trends across zones, time slots, etc.