Uber Interview Questions — Page 13

Uber·Software Engineer·Technical Phone Screen

Apr 2026

Uber SWE coding round, one problem the whole session. It looked like a jump game variant but had this extra wrinkle with digit-based jumps that took me a minute to even parse correctly.

You're given an integer array. Starting at index 0, you need to reach the last index. At each index i, you can move to i+1 or jump directly to any index j greater than i where the value at j ends in the digit 3. Your score is the sum of values at all visited indices. Find the maximum possible score.

“My first instinct was greedy and it was wrong.”

View Post

Uber·Software Engineer·Technical Phone Screen

Apr 2026

Uber SWE interview with a classic OOP design problem. Not as straightforward as it sounds once you get into the spot assignment logic and complexity analysis.

Design and implement a parking lot system with park, unpark, and checkcar methods. Motorcycle spots fit only motorcycles; regular spots fit motorcycles or cars. Define your class structure (ParkingLot, Spot, Vehicle types) and analyze the time complexity of each method.

“I started with the classes fine but the spot assignment logic tripped me up more than I expected.”

View Post

Uber·Software Engineer·Technical Phone Screen

Apr 2026

Uber SWE interview that came down to a single coding problem on quad trees. Not a topic I'd touched in a while so the first few minutes were a bit rough getting my bearings, but the structure of the problem was clean once I slowed down.

Implement a compressed QuadTree for an n x n grid where n is a power of two. Each region should be represented as a leaf node if all cells share the same value, otherwise split into four quadrants recursively. Build and return the root node.

“I fumbled the first few minutes because I kept second-guessing whether to check uniformity before or after recursing.”

View Post

Uber·Product Manager·Onsite - Product Sense / Strategy

Apr 2026

PM interview at Uber with a product design question about launching a kids-specific ride service. Pretty open-ended, no obvious right answer, which made it both interesting and kind of stressful.

Should Uber launch a ride service specifically for children? If yes, how would you design it?

“The 'should we' part tripped me up more than the design part.”

View Post

Uber·Data Scientist·Technical Phone Screen

Apr 2026

Uber Data Scientist coding screen, one problem the whole time: implement square root rounded to one decimal without using any built-in. Felt manageable but the edge cases and optimization discussion caught me a bit flat-footed.

Without using a built-in square root function, implement a function that returns the square root of a non-negative real number rounded to one decimal place.

“I went straight to binary search on the integer part and then iterated tenths, which worked but I fumbled when they asked about convergence and whether I could do better.”

View Post

Uber·Product Manager·Onsite - Product Sense / Strategy

Apr 2026

Product sense round at Uber focused on launching a new vertical within a super-app context. Just the one question but it had a lot of surface area to cover.

For a brand new vertical like peer-to-peer car rental inside a super-app, what metrics would you track in the first six months to evaluate whether it has legs?

“I went straight to supply and demand balance metrics, which felt right but I spent too long there and barely touched trust signals like damage claims or dispute rates.”

View Post

Uber·Software Engineer·Technical Phone Screen

Apr 2026

Uber follow-up round that built on a previous connectivity problem, this time throwing in block constraints that completely break the standard Union-Find approach. Harder than I expected and the complexity discussion at the end got uncomfortable fast.

You have an event log with ride-share events and block events between riders. A block means two riders can never be considered connected, even indirectly. Design an algorithm that finds the earliest timestamp at which all riders are connected without violating any block constraint, returning -1 if it's impossible.
Compare the time complexity of your BFS/DFS approach against the Union-Find baseline, and discuss whether offline processing, link-cut trees, or batched recomputation could improve it.

“This is a follow-up to a rider connectivity problem so they expected you to already know the baseline Union-Find solution.”

View Post

Uber·Data Scientist·Technical Phone Screen

Apr 2026

Uber DS interview, one big open-ended experiment design question that basically asked you to build an A/B test from scratch for a UI change. The kind of question where you realize halfway through that you forgot to think about interference effects and have to backtrack.

Uber is thinking about redesigning the pickup ETA card shown to riders after a trip request. The idea is that a clearer ETA display reduces cancellations. Design the full experiment end-to-end, covering randomization unit, target population, metrics, instrumentation, sample size, ramp strategy, bias controls, and success criteria.

“This one is a lot.”

View Post

Uber·Machine Learning Engineer·Technical Phone Screen

Apr 2026

Uber ML engineer interview, technical phone screen that was basically one meaty coding question about implementing multi-head self-attention from scratch in PyTorch, plus a bunch of follow-up theory questions. No shortcuts allowed, they explicitly said no nn.MultiheadAttention. Felt like a grad school exam.

Implement a MultiHeadSelfAttention module in PyTorch from scratch, without using any built-in attention layers. The class should accept embed_dim and num_heads, and the forward pass should take a tensor of shape (batch, seq_len, embed_dim) plus an optional mask.
Why do we scale the dot products by the square root of the head dimension before applying softmax?
What's the difference between causal attention and bidirectional attention, and how would you implement a causal mask?
Walk through the shape of the tensor at each step of the multi-head attention computation.
What is the time and memory complexity of self-attention with respect to sequence length?

“This is the kind of question where you think you know it until you're actually writing the code.”

View Post

Uber·Machine Learning Engineer·Technical Phone Screen

Apr 2026

Interviewed for an MLE role at Uber and got a pretty deep ML theory question about implementing the CLIP contrastive loss from scratch. Not a vibe check, they actually wanted working code and a real explanation of the design choices.

Given a batch of paired image and text embeddings, implement the symmetric contrastive loss used in CLIP-style training. You need to compute a full similarity matrix, then compute cross-entropy loss in both directions (image-to-text and text-to-image) and average them. Also explain your choices around normalization, temperature, and how you construct the labels.

“I knew the high-level idea of CLIP but writing it out under pressure was a different story.”

View Post