We use cookies for site analytics. Accept to help us understand how the site is used. See our Privacy Policy for details.
Prep for Uber's ML loop: marketplace ML problems, production ML systems, and coding.
Uber runs ML across its marketplace - ETA prediction, pricing, matching, fraud detection, and recommendations for Uber Eats - on an internal ML platform it has written about publicly (Michelangelo). ML engineer loops typically include coding, ML fundamentals, and ML system design for real-time marketplace problems, where latency, freshness of features, and feedback loops matter. Expect questions about evaluating models whose predictions change user and driver behavior, and about running models reliably in production. A behavioral round covers ownership and collaboration.
Regression and classification for ETA, pricing, and fraud; evaluation and drift.
Coding rounds, sometimes with a geo or scheduling flavor.
Real-time serving and feature pipelines.
Marketplace experiments have interference effects worth discussing.
Evaluation and uncertainty.
Ownership and collaboration stories.
Curated walkthroughs for the bounded designs that show up in Uber's system design rounds. Capacity estimation, architecture, deep-dives, and trade-offs.
Geo-indexing, real-time matching, ETA prediction, and surge. The canonical geo-spatial design problem with hard real-time constraints.
Online vs batch inference, GPU utilization tricks, autoscaling for spiky load, A/B testing models, and the feature store that decouples training from serving.
The offline/online store split, train/serve skew as the core problem it exists to solve, point-in-time-correct joins, materialization and freshness, and a registry that lets teams reuse features instead of re-deriving them badly.
The two-stage candidate-generation-then-ranking architecture, embeddings + ANN retrieval, the batch/real-time feature split, feedback loops, cold start, and why offline NDCG lies until the online A/B disagrees.
Sample STAR answers, common prompts, pitfalls, and follow-up strategies for the behavioral themes that decide Uber's loop.
Uber's systems move people and money in the physical world, and 'we act like owners' means unowned problems are yours the moment you see them. Interviewers test whether your ownership survives contact with someone else's code at 2am.
Uber has three customers on every trip - rider, earner, merchant - and a fix for one can quietly tax the others. Interviewers test whether your customer empathy survives multi-sided tradeoffs and global scale.
Tested at every level, scored harder at senior. Did you take responsibility for outcomes - or just for tasks?
Leaders operate at all levels. The interviewer is testing whether you actually understand your own systems - or whether you summarize what your team built.
The ml engineer learning path and the judgment cases that mirror this loop's execution and metrics questions.
About 50 hours, sequenced for Machine Learning Engineer loops.
Tests whether you understand liquidity as the marketplace health metric - measured per side, per market cell - and can find the binding constraint.
The most common analytical PM case. They want a structured, hypothesis-driven investigation - not a guess.
Total comp ranges, base, equity, and bonus across the levels tested in this loop. Aggregated from public sources.
6 ML engineer levels covered. Updated 2026-05.
348 MCQs and 175 coding challenges, grouped by topic. Free preview shows question titles - premium unlocks full content.
Behavioral and system design rounds reward practice with a live AI interviewer that probes follow-ups, not silent reading.
Start an AI mock interview →Marketplace problems are common: ETA prediction, pricing, matching, fraud detection, and Eats recommendations.
Riders and drivers share a pool, so a treatment can affect the control group. Mentioning interference and approaches like switchback or region-level tests shows depth.
No, but reading Uber's public engineering posts about its ML platform gives useful vocabulary for production ML design.