Uber

Role: SDE-ML2

College: Not Specified

Interview Date: 2025-09-05

Interview Type: Full-Time

Result: Probably Rejected ("Didn't make it")

Difficulty: Medium to Hard

Rounds: 5 rounds (Phone Screen, ML-Coding, ML-Design, DSA, Hiring Manager)

Topics Asked: Data Structures, Algorithms, Machine Learning Coding, Machine Learning System Design (Recommendation Systems), Behavioral

Experience:

The candidate, with a DS/ML background from a startup, applied for an SDE-ML2 position at Uber.

ACT-1 (Phone Screen): The candidate had 3 weeks to prepare. The question was a LeetCode Medium problem: weighted random number generation. The candidate explained optimal solutions and struggled a bit with coding but got it running. Result: Selected.

ACT-2 (Actual Rounds):

2.1: ML-Coding: The task was to take a fixed-length time series, perform a pairwise rank transformation, then use correlation as a distance metric to bucketize the time series into K clusters, with each cluster having the same number of elements. This felt like a Hard question. The candidate struggled with understanding the problem and ran out of time to code fully.

2.2: ML-Design: The task was to design a recommendation system for Uber Eats, focusing on new users. The candidate proposed solutions using popularity features and area-level user features. The explanation was not clear.

2.3: DSA: The question was a LeetCode Medium problem. Given a 2D grid of walls, robots, and pass-through blocks, plus a query (l,r,t,b), find all robot positions that satisfy the query. The candidate started with a brute force approach and arrived at the optimal solution within time.

Hiring Manager Round: Details not fully specified in the snippet, but it was a part of the main rounds.

The candidate hadn't heard back in 2 weeks, leading to the conclusion of not being selected. Despite not getting the offer, the candidate was proud of the progress made.