Salesforce·Software Engineer·Onsite - System Design / Architecture·Senior

Senior

Jun 2026

Summary

System design round at Salesforce for a software engineering role, focused entirely on building a movie recommendation system from scratch. Pretty broad scope for a single question but they clearly wanted to see how you'd structure ambiguity into something real.

Questions Asked (1)

Design a movie recommendation system end to end, covering everything from goal setting and data signals to modeling, evaluation, and rollout.

System DesignA/B Testing & ExperimentationProduct Analytics & Metrics

Author's notes

This one sprawls fast if you're not careful.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the business goal and constraints (e.g., increase watch time, user engagement, or retention). Then walk through the ML lifecycle: data collection, feature engineering, model selection, offline/online evaluation, and deployment with A/B testing. Emphasize trade-offs and how you would iterate based on metrics.

Pro tip: Tie every technical decision back to a measurable business metric and mention how you would handle cold-start and scalability, as these are common follow-ups. Also, discuss how you would incorporate user feedback loops and monitor for model drift.

1. Define Goals and Metrics

Clarify the objective (e.g., increase user engagement, watch time, or retention) and define success metrics (e.g., CTR, watch time per user, diversity). Consider business constraints and user experience.

2. Data and Signals

Identify data sources: user interactions (views, ratings, likes), item metadata (genre, cast, director), and contextual data (time, device). Discuss data preprocessing, feature engineering, and handling implicit/explicit feedback.

3. Modeling Approach

Choose models: collaborative filtering (matrix factorization, neural networks), content-based, or hybrid. Discuss cold-start strategies (e.g., using metadata or popularity) and scalability (e.g., distributed training, serving).

4. Evaluation and Offline Testing

Use offline metrics (RMSE, precision@k, recall@k, NDCG) and validate with holdout sets. Consider diversity, novelty, and serendipity. Simulate online performance with counterfactual evaluation.

5. Deployment and Online Experimentation

Deploy via A/B testing, measure impact on business metrics, and iterate. Discuss infrastructure (real-time serving, caching), monitoring, and feedback loops for continuous improvement.

Key Points to Mention

Cold-start problem and solutions (e.g., using metadata, popularity, or contextual bandits)
Hybrid recommendation approach combining collaborative filtering and content-based methods
Offline vs. online evaluation metrics and their limitations
A/B testing framework, including guardrail metrics and statistical significance
Scalability considerations: distributed training, real-time inference, and latency
Feedback loops and model retraining to handle concept drift

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