Upstart Interview Questions — Page 6

Upstart·Data Scientist·Technical Phone Screen

May 2026

Upstart data science interview threw a pretty heavy optimization problem at me, the kind where you need to know your constrained regression theory cold. It was a technical screen that felt more like a quant research exercise than a typical DS interview.

Given a 65-element vector of yearly discretionary income, formulate a spending plan that is nondecreasing over time, never borrows from future income, and minimizes the squared deviation from the income profile. Show that this reduces to isotonic L2 regression under a nondecreasing constraint, and explain why the no-borrowing condition is automatically satisfied when probability mass is only shifted forward.
Implement an O(n) Pool-Adjacent-Violators Algorithm in Python or R that returns both the smoothed spending vector and the block structure of the solution.
Using a fixed random seed and a normally distributed income vector (mean 1000, sd 100, clipped at 0), run your algorithm and report the total squared error, verify the output is nondecreasing, and confirm that cumulative spending never exceeds cumulative income at any time step.
Compare your PAVA implementation against a naive iterative averaging approach (repeatedly smooth out decreasing runs) in terms of both correctness and runtime.

“This is where I spent most of my mental energy.”

View Post

Upstart·Data Scientist·Technical Phone Screen

May 2026

A/B testing deep dive for a Data Scientist role at Upstart. The question was essentially a full experiment design case built around a Pinterest-style video feed change, and it covered a lot of ground in one shot.

You're a data scientist at a platform like Pinterest. The team wants to increase the share of video content in the home feed to drive more engagement. Design the A/B test end-to-end, including hypotheses, experiment unit, metrics, segmentation, rollout plan, and stopping rules.

“This question is basically five questions dressed up as one.”

View Post

Upstart·Software Engineer·Technical Phone Screen

May 2026

Interviewed for a software engineering role at Upstart and got a string filtering problem that looked simple on the surface but had a few layers worth thinking through.

Given a list of strings and a reference string s, return only the strings from the list that share at least one character with s. What's an efficient way to implement this, and how do you handle case sensitivity and Unicode?

“The core solution came pretty naturally.”

View Post

Upstart·Data Scientist·Technical Phone Screen

May 2026

Upstart data scientist interview that went pretty deep into probability theory and simulation, way more math-heavy than I expected for a DS role. The drunk passenger problem was the centerpiece and they wanted everything: closed-form proof, generalization, working code, and a complexity discussion.

There are 100 passengers and 100 seats. Passenger 1 picks a random seat. Each subsequent passenger sits in their assigned seat if it's free, otherwise picks randomly from what's left. Derive a closed-form expression for the probability that passenger 100 ends up in seat 100.
Generalize your result to n passengers and n seats (n >= 2) and give a brief proof, either by induction or by identifying an invariant.
Write a Monte Carlo simulator in Python or R to estimate this probability for n=100 using at least 1 million trials. Report the estimate, standard error, and a 95% confidence interval, and confirm it matches theory within 0.01.
What are the time and space complexities of a naive simulation versus an optimized approach that only tracks the two boundary seats (seat 1 and seat n)?

“The classic drunk passenger problem.”

View Post

Upstart·Software Engineer·Technical Phone Screen

May 2026

Interviewed for a software engineer role at Upstart and got a string manipulation problem that looked deceptively simple on the surface. The anagram matching with a tiebreaker condition was the kind of thing that trips you up if you rush straight into code.

Given a target string and a list of candidate strings, find which candidate is an anagram of the target. If more than one qualifies, return the one whose first character matches the first character of the target. Return an empty result if nothing matches.

“My first instinct was to just sort both strings and compare, which works fine for the anagram check.”

View Post

Upstart·Data Scientist·Technical Phone Screen

May 2026

Data science interview at Upstart centered on a pretty meaty A/B testing scenario involving two treatments and conflicting metric signals. The whole thing felt like one long case question with follow-ups that kept getting harder.

Treatment t1 shows no significant change in either Gross Booking or Variable Consideration, while t2 shows a significant GB increase but a significant VC decrease. How do you explain these results to a PM and what do you recommend doing next?
Given that t2 has a GB confidence interval of [+0.1%, +2.3%] with a +$0.48 lift and a VC confidence interval of [-2.5%, -1.5%] with a -$0.20 loss, would you launch it? Walk through your reasoning.
How would you segment users or orders to find cohorts where t2 drives positive GB impact without the negative VC impact?
If 20 experiments are running at the same time, how do you set launch criteria that control for false discoveries and still let you make reliable decisions?

“I spent too long on the mechanics of the stats and not enough on the story for the PM.”

View Post

Upstart·Data Scientist·Onsite - System Design / Architecture

May 2026

Upstart data scientist interview that went pretty deep into ML system design for ads. The question was broad enough that you could spend an hour on any single piece of it, which made scoping feel like half the challenge.

Design a complete machine learning system for ad selection on a platform serving millions of users in real time. Walk through data collection, labeling, feature engineering, model choice, exploration vs. exploitation, online and offline evaluation, latency constraints, and scalability.

“This is a monster of a question.”

View Post

Upstart·Data Scientist·Onsite - System Design / Architecture

May 2026

System design round at Upstart for a data scientist role. The whole thing was one big open-ended question about ads serving, and they really wanted you to go deep on every layer of the stack. Felt like a senior-level bar even if the title didn't say so.

Walk through the end-to-end design of an ads serving system for a social media app, covering data collection, feature engineering, model selection, real-time ranking, feedback loops, and A/B evaluation. Also discuss offline vs online metrics and how you'd handle cold-start users.

“This is basically a dissertation prompt.”

View Post

Upstart·Data Scientist·Technical Phone Screen

Jun 2026

Big technical case for a DS role at Upstart. The whole question was structured around a NerdWallet-style comparison page and it covered basically everything: instrumentation, metrics, experimentation, insight generation, cannibalization. A lot to hold in your head at once.

How would you instrument a loan comparison page and both its on-site and click-out application flows end-to-end, including event naming, join keys, cross-domain tracking, and session stitching when cookies are unreliable?
Define success metrics and guardrails for both the comparison page and each individual partner, with precise numerator and denominator definitions for things like CTR, on-site completion rate, qualified rate, funded loan rate, and revenue per session.
Looking at visible differences in partner listings such as APR ranges, fees, and eligibility messaging, generate three non-obvious insights about how those differences might shift user mix and downstream funding rates.
Propose an experiment comparing on-site versus click-out application entry points for a given partner, covering experimental unit, randomization, stratification, sample size and power assumptions, a pre-registered decision rule, and how you'd handle selection bias and attribution mismatches.
How would you measure partner cannibalization and cross-listing traffic contamination, accounting for things like users revisiting the page, tab hoarding, and last-click attribution bias?

“This is the part I felt least confident about.”

View Post

Upstart·Software Engineer·Technical Phone Screen

Jun 2026

Did a technical screen for a software engineer role at Upstart. Just one coding problem, pretty short session, nothing too crazy but it required more careful thinking than I expected.

Write a function that takes an integer representing a number of bytes and returns a formatted string showing the value in the appropriate unit (B, KB, or MB) with two decimal places. 1 KB = 1024 B, 1 MB = 1024 KB.

“Seemed straightforward but I fumbled the boundary conditions at first.”

View Post