DoorDash·Machine Learning Engineer·Onsite - System Design / Architecture·Senior

Senior

Apr 2026

Summary

DoorDash ML engineer interview with a system design question around building a discount/promotions system. Pretty open-ended, which I wasn't fully prepared for.

Questions Asked (1)

Design a system that provides discounts to customers.

System DesignProduct StrategyTechnical Trade-offs

Author's notes

I started with the ML angle immediately and probably should've scoped the problem more before jumping in.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the business objective and constraints, then propose a high-level system design that integrates ML models for discount optimization with a scalable serving infrastructure. Emphasize trade-offs between personalization, business rules, and real-time performance, and discuss how you would measure success and iterate.

Pro tip: Frame the problem as a constrained optimization: maximize long-term customer value (e.g., retention, order frequency) while respecting budget and fairness constraints. This shows you think beyond model accuracy and consider business impact.

1. Clarify Requirements and Objectives

Ask questions to understand the goal: Is the discount to acquire new customers, increase order frequency, or clear inventory? Identify constraints like budget, fairness, and latency.

2. Define Success Metrics and Data

Propose metrics such as incremental profit, conversion rate, and customer lifetime value. Discuss data sources: user behavior, order history, and contextual signals.

3. Design ML Model and System Architecture

Outline a model (e.g., uplift modeling or reinforcement learning) to predict discount response. Describe a scalable architecture with offline training and online serving, including feature store and A/B testing.

4. Address Trade-offs and Constraints

Discuss trade-offs: personalization vs. simplicity, real-time vs. batch, and exploration vs. exploitation. Explain how to handle budget pacing and fairness.

5. Plan Evaluation and Iteration

Describe offline evaluation (e.g., counterfactual estimation) and online A/B tests. Outline monitoring, feedback loops, and continuous improvement.

Key Points to Mention

Uplift modeling or causal inference to estimate incremental impact of discounts
Multi-armed bandits or reinforcement learning for exploration-exploitation trade-off
Real-time serving infrastructure with low latency and high scalability
Budget constraints and pacing algorithms to avoid overspending
Fairness and anti-discrimination considerations in discount allocation
A/B testing framework and guardrail metrics to measure long-term effects

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.