← Back to Directory

Upstart interview notes, page 1 of 10

Mid-sized

Most of these Upstart write-ups come from Software Engineers (51), then Data Scientists (42) and Machine Learning Engineers (2). The most reported rounds are the technical phone screen (73), the behavioral / leadership round (8), the online assessment (4) and the multi-round onsite (4).

97 interview notes · updated Jul 2026

Upstart·Software Engineer·Technical Phone Screen

Apr 2026
Straightforward coding screen for a software engineer role at Upstart. One question, pretty short session, nothing too wild.
  • Given a list of integers, return a reversed version of the list with all odd numbers removed.

“Easier than I expected.”

View Post

Upstart·Software Engineer·Technical Phone Screen

Apr 2026
Interviewed for a Research Scientist role at Upstart and got hit with a probability puzzle that felt deceptively clean but had a lot of moving parts underneath. The kind of question where you think you know where it's going and then realize halfway through you've been sloppy with your assumptions.
  • In a coin-guessing game where a player gets banned after 10 consecutive correct guesses, and each player has at most 100 guesses, derive an analytical expression for the expected number of banned players out of a population of N. Also discuss the probability of hitting a streak of 10 within 100 trials and any approximations you use.

“I started with the easy part: each guess is correct with probability 0.5, so a streak of 10 has base probability (0.5)^10.”

View Post

Upstart·Software Engineer·Technical Phone Screen

Apr 2026
Coding round at Upstart for a software engineer role. The problem looked like a cipher/crypto puzzle at first glance but was really just a composition of two substitution maps, which made it more approachable once I saw what was actually being asked.
  • You have two substitution cipher mappings: one used between sender A and B, and another between B and C. Given the final ciphertext that D receives after both substitutions have been applied, recover the original plaintext A wrote.

“My first instinct was to try reversing each map separately in sequence, which would've worked but felt clunky.”

View Post

Upstart·Product Manager·Onsite - Behavioral / Leadership

Apr 2026
Upstart had me do a third project deep-dive, which honestly felt like they were stress-testing whether I had range beyond the first two stories I'd already told.
  • Walk us through a third project that showcases something different from your previous two examples, such as leadership under ambiguity, cross-team collaboration, or a distinct technical area. Cover the problem, your specific contribution, the technical depth involved, and the concrete outcome.

“Picking the right project tripped me up more than answering it.”

View Post

Upstart·Software Engineer·Technical Phone Screen

Apr 2026
Coding round at Upstart for a software engineer role. Pretty focused on arrays and sorting, nothing too wild, but they pushed on complexity tradeoffs which I wasn't fully ready for.
  • Given an array of numbers, find the minimum absolute difference between any two elements in the array.

“I knew sorting was the move but fumbled explaining why adjacent elements are the only pairs you need to check after sorting.”

View Post

Upstart·Machine Learning Engineer·Technical Phone Screen

Apr 2026
Interviewed for an ML Engineer role at Upstart and got hit with a pretty deep question on logistic regression loss functions. Not a casual screen by any means.
  • What is the loss function used in logistic regression? Walk through the binary cross-entropy expression, derive it from a probabilistic perspective using maximum likelihood, and explain why it's used instead of mean squared error for classification tasks. Bonus if you can extend the idea to the multi-class case.

“This one went deeper than I expected for a phone screen.”

View Post

Upstart·Software Engineer·Technical Phone Screen

Apr 2026
Had a technical phone screen for a software engineer role at Upstart. One algorithmic problem, fairly well-defined but with a sneaky constraint that makes it less obvious than it looks at first glance.
  • You're given a list of cities and their longitudes. Traveling west to east, find the maximum number of cities you can visit such that each subsequent city is both further east (higher longitude) and lexicographically greater in name than the previous one.

“The longitude constraint alone would just be longest increasing subsequence, which I spotted pretty fast.”

View Post

Upstart·Data Scientist·Onsite - Multi Round

Apr 2026
Onsite ML case at Upstart for a Data Scientist role. The session was focused on a single meaty problem about missing training data and how you'd still ship a model that generalizes. Not a lot of back-and-forth, more like a design session where they watched you think out loud.
  • You're working with California housing data but the training set is missing the lowest income bracket (under $25k). How do you build a model that still generalizes well to that unseen group?

“This one took me a second to even parse correctly.”

View Post

Upstart·Software Engineer·Technical Phone Screen

Apr 2026
Got a coding question at Upstart for a software engineer role that was more interesting than your typical string manipulation problem. The anagram decoding twist with a shared first/last character constraint made it a step above the usual.
  • Given a vocabulary list and a string of scrambled words, decode the string. Each scrambled word is an anagram of exactly one vocabulary word and shares the same first and last character with it. Return the decoded string.

“My first instinct was to just sort the characters of each word and use that as a lookup key, which works until it doesn't.”

View Post

Upstart·Software Engineer·Technical Phone Screen

Apr 2026
Went through a coding round at Upstart for a QA Engineer role and got hit with what felt like a classic SWE problem. Not the most QA-flavored interview I've had, but it was fine.
  • Given an array of integers and a target value, return the indices of the two numbers that sum to the target. Walk through both a brute-force approach and an optimized solution, and discuss follow-up variants like handling duplicates, sorted arrays, or counting all valid pairs.

“Started with the nested loop approach just to show I understood the baseline, then moved to the hash map version where you check the complement on each pass.”

View Post