Build a package-allocation model for couriers
Overview
This question evaluates a candidate's ability to design an end-to-end ML-driven package-to-courier allocation system, testing competencies in predictive modeling (per-stop service-time forecasting), constrained combinatorial optimization for routing and fairness, feature engineering, simulation, and production concerns such as cold-start handling and distribution shift. It is asked to assess how candidates balance competing objectives like on-time delivery versus fairness under real-world constraints (shift hours, vehicle capacity, delivery windows, time-varying travel forecasts), and is categorized under Machine Learning with a focus on practical application-level system design rather than purely conceptual theory.
Automatic Package-to-Courier Assignment with ML + Optimization
You previously assigned packages to couriers manually. Design an end-to-end system that automatically assigns packages per courier per shift, optimizing for both on-time delivery and fairness.
Provide the following:
- Objective Function(s)
- Define a primary objective (e.g., maximize on-time probability − λ·overtime − μ·distance) and fairness criteria (e.g., limit variance of workload or use CVaR fairness).
- Constraints
- Include: shift hours, vehicle capacity (weight/volume), delivery windows, geographic contiguity, skill requirements (e.g., hazmat), and realistic service-time behavior (from historical stop-time distributions).
- Required Data Features
- Specify features such as route density, travel-time forecasts by hour and weather, package size class, stop difficulty, and historical courier productivity.
- Algorithmic Approach
- Predict per-stop service time using ML (e.g., gradient boosting with prediction intervals) and feed predictions into a constrained optimizer (integer/linear programming, VRPTW, min-cost flow, or column generation).
- Explain how you will handle cold-start couriers, guard against distribution shift, and design a fast rebalancing heuristic for last-minute spikes.
- Pseudocode
- Provide pseudocode for the end-to-end loop: forecast → optimize → simulate → assign → monitor.
- Explain fallback behavior if the optimizer fails or times out.
Solution
Login required