← Tubitv Interview Insights

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

SeniorPrefer not to say
Jul 2026Remote

Summary

ML system design round at Tubitv for an MLE role, focused entirely on building a thumbnail selection system for a streaming catalog. Pretty deep dive, lots of follow-ups that pushed into territory I wasn't fully prepared for.

Questions Asked (5)

Q1

Design an end-to-end ML system that selects which thumbnail to show each user for a given streaming title, with the goal of maximizing engagement. Cover problem framing, data, modeling, offline and online evaluation, serving, and production monitoring.

System DesignTechnical Trade-offsProduct Analytics & Metrics
Author's notes

This was the main question and it ate up most of the session.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by framing the problem as a contextual bandit or ranking task where the goal is to maximize user engagement (e.g., clicks or watch time) by selecting the best thumbnail for each user-title pair. Walk through the full ML lifecycle: data collection and feature engineering, model training and offline evaluation, online A/B testing, scalable serving, and monitoring for drift and feedback loops. Emphasize trade-offs between personalization and exploration, and how to handle cold-start and position bias.

Pro tip: Highlight the importance of logging propensities and using off-policy evaluation to safely test new models offline, and mention that thumbnail selection is a classic exploration-exploitation problem where a small percentage of traffic should always be reserved for exploration to avoid feedback loops.

1. Problem Framing and Metrics

Define the task as a ranking or classification problem: given a user and a title, predict engagement for each candidate thumbnail. Choose primary metrics (e.g., click-through rate, watch time) and guardrail metrics (e.g., user satisfaction, diversity).

2. Data and Feature Engineering

Collect historical interaction data (impressions, clicks, watch time) with user, title, and thumbnail features. Engineer features such as user demographics, viewing history, title genre, thumbnail visual features (via CNN embeddings), and context (time of day, device).

3. Modeling and Offline Evaluation

Train a model (e.g., gradient boosted trees or deep neural network) to predict engagement probability. Use offline evaluation with counterfactual methods (e.g., inverse propensity scoring) and simulate A/B tests to estimate potential uplift.

4. Online Evaluation and Serving

Deploy via A/B test with a small exploration bucket. Serve predictions in real-time using a low-latency model server, with fallback to popularity-based thumbnails. Log all decisions and outcomes for continuous learning.

5. Production Monitoring and Iteration

Monitor model performance, data drift, and business metrics. Set up alerts for degradation. Periodically retrain with new data and conduct offline-online correlation checks to ensure offline metrics reflect online impact.

Key Points to Mention

  • Contextual bandit formulation to balance exploration and exploitation
  • Handling position bias and using inverse propensity scoring for unbiased offline evaluation
  • Feature engineering for thumbnails: visual embeddings, text, and user-title interaction features
  • Cold-start strategies for new users and new titles (e.g., content-based fallbacks)
  • Real-time serving architecture with low latency and scalability
  • Monitoring for feedback loops, data drift, and metric degradation

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

Q2

Your training data only has feedback for thumbnails that were actually shown. How do you reason about thumbnails that were never shown to a given user, and how do you evaluate a model that needs to compare all candidates?

A/B Testing & ExperimentationTechnical Trade-offsAlgorithms & Data Structures
Author's notes

This one genuinely stumped me for a moment.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Acknowledge the missing-not-at-random problem and propose a counterfactual or off-policy evaluation framework. Explain how to use techniques like inverse propensity scoring, doubly robust estimation, or model-based imputation to estimate performance on unshown thumbnails. Emphasize the need for a full-ranking evaluation metric and possibly a simulation or offline A/B test to validate.

Pro tip: Mention that you would log propensities and use them to correct for bias, and that you would validate with a small randomized exploration bucket to get unbiased ground truth. This shows you understand both the theory and the practical need for experimentation.

1. Define the problem and assumptions

Clarify that the training data is biased because thumbnails are shown based on a previous policy, and unshown thumbnails have no feedback. State assumptions about the logging policy and whether it's known or can be estimated.

2. Choose an off-policy evaluation method

Select a technique such as inverse propensity scoring (IPS), doubly robust (DR), or model-based imputation to estimate the performance of a new policy on all candidates. Discuss trade-offs between variance and bias.

3. Design the evaluation metric

Propose a full-ranking metric like NDCG or mean reciprocal rank that considers all candidates, and explain how to compute it using the off-policy estimates. Address how to handle missing data for unshown items.

4. Validate with a small randomized experiment

Suggest running a small A/B test where thumbnails are randomly assigned to users to collect unbiased feedback. Use this to calibrate and validate the off-policy estimates.

5. Iterate and monitor

Emphasize the importance of continuous monitoring and re-evaluation as the policy changes, and the need to update propensities and models over time.

Key Points to Mention

  • Missing not at random (MNAR) and selection bias in logged feedback
  • Inverse propensity scoring (IPS) and its variance issues
  • Doubly robust estimation combining IPS with a reward model
  • Counterfactual reasoning and off-policy evaluation
  • Full-ranking metrics like NDCG for comparing all candidates
  • The value of a small randomized exploration bucket for unbiased validation

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

Q3

If a click-maximizing model starts surfacing sensational or slightly misleading thumbnails that get clicks but result in low watch completion, how do you detect that and adjust the objective?

Product Analytics & MetricsRoot Cause AnalysisTechnical Trade-offs
Author's notes

I actually liked this one.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by acknowledging the trade-off between click-through rate (CTR) and watch completion, then propose a multi-metric monitoring system to detect divergence. Explain how to adjust the objective function to incorporate watch completion or other engagement metrics, and validate the change through A/B testing.

Pro tip: Emphasize the importance of aligning the objective with long-term user satisfaction and retention, not just short-term clicks. Mention that misleading thumbnails can erode trust and lead to churn, so the fix should balance engagement with quality.

1. Detect the Issue

Monitor key metrics like CTR, watch completion rate, and user retention. Set up alerts for significant divergence between CTR and watch completion, indicating misleading thumbnails.

2. Analyze Root Cause

Investigate whether the drop in watch completion is due to sensational thumbnails by analyzing thumbnail content, user feedback, and session-level data. Conduct qualitative analysis or surveys to confirm.

3. Adjust Objective Function

Modify the model's objective to include watch completion or a weighted combination of CTR and watch time. Consider using a multi-task learning approach or a composite metric like CTR * watch completion.

4. Validate with A/B Test

Run an A/B test comparing the old click-maximizing model with the new objective. Measure impact on CTR, watch completion, and long-term metrics like retention and user satisfaction.

5. Monitor and Iterate

Continuously monitor the new model's performance and iterate on the objective weights or add additional quality signals to prevent future misalignment.

Key Points to Mention

  • Metric divergence detection (CTR vs. watch completion)
  • Root cause analysis using qualitative and quantitative data
  • Objective function design (multi-task learning, weighted metrics)
  • A/B testing methodology and long-term impact measurement
  • User trust and retention as ultimate goals
  • Iterative monitoring and adjustment

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

Q4

A brand-new title enters the catalog with five thumbnails and zero engagement history. Walk through exactly what the system does for that title over the first hours and days.

System DesignAdaptability & AmbiguityTechnical Trade-offs
Author's notes

Cold start.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Structure your answer as a timeline: immediate (first minutes), short-term (first hours), and medium-term (first days). For each phase, describe the system's actions, the ML models involved, and how the system adapts as engagement data arrives. Emphasize the cold-start problem and how you balance exploration with exploitation.

Pro tip: Show that you understand the business context: Tubi's success depends on quickly surfacing new content to the right users. Mention that you would instrument the system to collect feedback and use it to improve recommendations, while avoiding premature optimization.

1. Initial Ingestion and Feature Extraction

When the title is added, the system extracts metadata (genre, cast, director, etc.) and processes the five thumbnails using computer vision models to generate visual embeddings. These features are stored in a feature store for immediate use.

2. Cold-Start Recommendation Strategy

Since there is no engagement history, the system relies on content-based filtering: match the title's features to users with similar preferences. Use a multi-armed bandit approach to allocate some impressions to the new title for exploration, while exploiting known popular titles.

3. Real-Time Monitoring and Feedback Loop

As users interact (clicks, watches, likes), the system logs these events in real-time. A streaming pipeline updates user and item embeddings incrementally, and the bandit algorithm adjusts the exploration rate based on early performance.

4. Short-Term Model Adaptation

Within hours, the system retrains or fine-tunes the ranking model using the new data. It may also perform A/B tests on thumbnail selection to optimize click-through rate, using the visual embeddings to predict which thumbnail appeals to which user segment.

5. Medium-Term Optimization and Evaluation

Over days, as more engagement data accumulates, the system transitions from content-based to collaborative filtering signals. It evaluates the title's performance against baselines and decides whether to increase or decrease its promotion in recommendations.

Key Points to Mention

  • Cold-start problem and content-based filtering using metadata and visual features
  • Multi-armed bandit for exploration-exploitation trade-off
  • Real-time data pipeline for ingesting engagement events
  • Incremental learning or online updating of models
  • Thumbnail selection and A/B testing for optimizing CTR
  • Transition from content-based to collaborative filtering as data grows

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

Q5

How would you determine whether personalizing thumbnail selection per user is actually worth the added complexity compared to just picking one globally best thumbnail per title? What experiment would you run?

A/B Testing & ExperimentationProduct StrategyTechnical Trade-offs
Author's notes

Straightforward A/B framing: run a global-best variant against a personalized variant and measure engagement lift, then weigh that against infrastructure cost and latency.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Frame the decision as a cost-benefit analysis: quantify the potential lift from personalization against the engineering, infrastructure, and maintenance costs. Propose a rigorous A/B test that measures both user engagement metrics and long-term retention, while accounting for novelty effects and cold-start users.

Pro tip: Emphasize that personalization should only be pursued if the lift is substantial and sustainable, and if the infrastructure can support it without degrading latency or reliability. Suggest starting with a shadow deployment or small-scale test to estimate potential gains before full investment.

1. Define Success Metrics and Hypotheses

Identify primary metrics (e.g., click-through rate, watch time, retention) and secondary metrics (e.g., diversity, user satisfaction). Formulate a clear hypothesis: personalization will improve X by Y% for Z% of users.

2. Estimate Costs and Complexity

List the added complexity: data pipelines, model training/serving, real-time inference, monitoring, and potential technical debt. Estimate engineering effort and ongoing operational costs.

3. Design the Experiment

Propose an A/B test with a control (global best thumbnail) and treatment (personalized thumbnails). Ensure proper randomization, sufficient power, and guardrail metrics. Consider a switchback or holdout design to measure long-term effects.

4. Analyze Results and Make a Decision

Compare lift in primary metrics against costs. Check for heterogeneity across user segments (e.g., new vs. existing, heavy vs. light viewers). Decide if the lift justifies the complexity, and consider a phased rollout if promising.

5. Plan for Iteration and Monitoring

If personalization is adopted, set up continuous monitoring and periodic re-evaluation. If not, document learnings and revisit if costs decrease or personalization techniques improve.

Key Points to Mention

  • A/B testing methodology: randomization, sample size, statistical power, and avoiding peeking
  • Metrics: CTR, watch time, retention, and guardrail metrics like latency and diversity
  • Cost-benefit analysis: engineering effort, infrastructure costs, and maintenance overhead
  • Cold-start problem: how to personalize for new users or new content
  • Novelty effect: short-term lift may not persist; use long-term holdout or cohort analysis
  • Segmentation: personalization may benefit some user segments more than others

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