Goldman Sachs CodePad Interview: 8 Questions to Expect

The CodePad round is a live, 45-to-60-minute coding interview where a Goldman Sachs engineer watches you solve one to two problems in a shared editor, and your code has to actually run. It sits in the middle of the engineering funnel, after the HackerRank assessment and before Superday, and it cuts more candidates than either neighbor. Recent candidates describe the difficulty as one LeetCode easy to warm up, then one medium that decides the round. For some engineering recruiting processes, CoderPad comes after an online assessment and before later-stage interviews or Superday. Recent reports also show that the difficulty can vary considerably: some candidates have encountered two relatively straightforward problems, while others report medium-level questions involving dynamic programming or binary search.

Goldman opened 2027 Summer Analyst applications on August 15, 2026., so the students WSG coaches toward engineering seats are staring this round down right now. Here's the funnel, the eight most-reported questions, and what the interviewer's rubric rewards.

CodePad is just one stage in Goldman's process, walk through the HireVue round that usually comes with it here & see how a similarly rigorous prep looks at another elite bank here.

What is the Goldman Sachs CodePad interview?

CodePad is the name candidates and recruiters use for Goldman's live technical screen, run on a collaborative editor in the CoderPad style with an engineer present. It is a conversation, never an async test: you talk through your approach, code it, run it, and debug it while they watch. US campus candidates often meet it as two back-to-back 45-minute technical interviews.

Is CodePad the same as the HackerRank test?

No. Goldman's official prep page says engineering applicants take a HackerRank assessment, and that's the earlier, automated stage: timed problems, no human watching, formats ranging from two to four coding questions to mixed sets with logic and even machine-learning multiple choice. The exact HackerRank format can vary by role and recruiting cycle, so candidates should not assume every applicant receives the same number or type of questions.   CoderPad is a separate, live interview stage in which you solve coding problems while explaining your approach to a Goldman Sachs engineer. Recent candidate reports commonly describe two coding problems, though the number and difficulty can vary. The distinction matters for preparation: HackerRank rewards speed and independent problem-solving; CoderPad adds communication, debugging, and the ability to defend your approach under questioning. 

The funnel, end to end

The 2025-26 reported sequence for engineering campus hires: apply through Goldman's students hub, complete the HackerRank assessment, record the HireVue (five to seven questions, mostly behavioral with one technical and one markets-flavored), then the CodePad round or rounds, then a Superday of two to five interviews mixing coding, light system design, object-oriented concepts, and behaviorals. Average reported timeline runs about eight weeks application to offer. Formats drift by region and cohort, so treat recruiter emails as the authority on your version.

Two tracks, one funnel

Goldman splits campus engineering into Summer Analyst roles, the nine-to-ten-week internship for penultimate-year students, and New Analyst roles for final-year students and recent grads going straight to full time. Both tracks report the same stage sequence, and both route through the same application portal. Regional cohorts differ more than tracks do: India cycles opened in early July 2026 with the Americas following August 15, and reported assessment formats vary most across regions

Note which division you're applying to as well. Engineering sits alongside Goldman's banking and markets tracks, with its own recruiters and its own interview content. Nothing in this article covers the IB technical interview; a DCF will not save you in a CodePad round.

The 8 questions to expect

All eight below come from candidate interview reports on Glassdoor, LeetCode discussions, and engineering forums from 2021 through 2026. Goldman rotates specifics, so drill these as patterns rather than answers.

1. Trapping rain water

The single most-reported Goldman question, appearing in both CodePad and Superday reports. You’re given a list of heights representing walls. Calculate how much water can be trapped between the walls after it rains. The simplest solution takes O(n²) time, but the expected solution is O(n) using either prefix maximums or the two-pointer method. Interviewers may ask you to improve your first solution so that it uses less memory.

2. Compress a string with run-length encoding

A common Goldman Sachs string question: Compress a string by replacing repeated letters with the letter followed by how many times it appears. For example, "aaabbccaaaa" becomes "a3b2c2a4".  Be careful to include the last group of letters, since it’s easy to accidentally leave it out when the loop ends. A common follow-up is: What if the compressed version is longer than the original string? Would you still return it, or just return the original string instead?

3. Longest substring without repeating characters

A common Goldman Sachs coding pattern: Use a sliding window and a hashmap to solve string/array problems.

For example, when looking for the longest substring without repeated characters, keep track of where you last saw each character. When you see a repeat, move the left side of your window forward. Keep track of the longest window you find.

The goal is to be able to code this in about 10 minutes without looking at notes.

4. Minimum meeting rooms for overlapping intervals

You’re given a list of time intervals, such as meeting times, and need to figure out how many meetings can happen at the same time.

Sort the meetings by their start time, then use a min-heap to keep track of when each meeting ends. The largest size the heap reaches tells you the minimum number of rooms you need.

The key idea to remember is: sort by start time → track end times → count the maximum overlap.

5. Design an LRU cache

Build an LRU (Least Recently Used) cache, which keeps track of the items that were used most recently and removes the ones that haven’t been used in the longest time.

Use a hashmap to quickly find items and a doubly linked list to keep track of the order they were used. Together, they let you add, remove, and find items in O(1) time.

The important part isn’t just writing all the code. It’s being able to explain why you need both a hashmap and a linked list.

6. Group anagrams

You’re given a list of words and need to group words that contain the same letters. For example, "eat", "tea", and "ate" would go in the same group.

One way to do this is to sort the letters in each word and use the sorted word as the key in a hashmap. A faster option is to count how many times each letter appears and use those counts as the key.

A common follow-up is: What is the difference in time complexity between sorting the letters and counting them?

7. Shortest time to reach every node in a network

Imagine you have different machines connected to each other, and you want to figure out how long it takes for information to travel from one machine to all the others.

If every connection takes the same amount of time, use BFS (Breadth-First Search). Start at the first machine and work outward, keeping track of how many steps it takes to reach each machine.

A common follow-up is to make the connections take different amounts of time. In that case, you would use Dijkstra’s algorithm instead of BFS because a regular queue can’t handle different travel times correctly.

8. Valid parentheses

Check whether a string has matching brackets. For example, "({[]})" is valid, while "([)]" is not.

Use a stack: put opening brackets like (, {, and [ onto the stack. When you see a closing bracket, check that it matches the most recent opening bracket.

If everything matches and the stack is empty at the end, the string is valid.

The three mistakes that end rounds early

Debriefs from students we've coached through this loop repeat three failure patterns, and none of them is "couldn't solve it."

Coding before agreeing on the problem. Strong candidates restate the input, the output, and one edge case, then ask "is that the right read?" before writing a line. Weak candidates burn fifteen minutes building the wrong function beautifully.

Ignoring the run button. The editor executes code, and interviewers report that candidates who never run anything until the end signal that they don't work the way engineers work. Run early, run often, and narrate what each failure tells you.

Freezing on the optimization ask. "Can you do better than O(n squared)?" is an invitation, never an ambush. The expected response is thinking out loud about which data structure buys down the inner loop. Silence is the only wrong answer.

The HireVue and Superday bookends

The HireVue deserves an evening of its own prep. Recent engineering candidates report seven questions in about 30 minutes, five behavioral, one technical, one finance-adjacent, each with 30 seconds of prep and two minutes of answer. Practice recording yourself against a timer: the format rewards answers that are structured and concise, particularly when you have limited preparation time.

Superday is the final stretch, but the format varies significantly by team and role. Superday runs two to five interviews in a day, and reports from early 2026 describe one round of pure problem solving and one mixing system design at survey depth, object-oriented questions like singleton and factory patterns, and behaviorals with your resume open. Teams vary: a strat-leaning desk asks probability, a platform team asks about services and caching. Superday interviewers assume you can code by that stage, so the differentiator becomes whether they'd want to sit next to you during an outage.

What the interviewer is actually grading

Engineers who run these rounds describe the rubric consistently: state an approach before coding, write code that compiles and runs, name the time and space complexity unprompted, and test your own edge cases before declaring victory. One Superday report put it plainly, that they want quality code rather than a math wizard. Communication is scored the whole way through, which is why silent perfect solutions underperform talkative imperfect ones.

Behavioral threads run through technical rounds too. The HireVue mixes in questions like how you'd debug a production issue and one finance-adjacent prompt such as how you'd value a company, at survey depth only.

How to prepare in four weeks

Week one: 20 easy problems on arrays, strings, and hashmaps, all typed into a bare editor without autocomplete, since the CodePad environment strips your IDE comforts. Weeks two and three: 25 mediums weighted toward the patterns above, plus HackerRank's interview kit to match the OA's house style. Week four: mock interviews out loud, ideally with a CS friend playing a probing interviewer, because narrating while coding is a separate skill from coding.

Pick one language and stay in it; candidates report Python, Java, and C++ all land fine, and interviewers occasionally quiz fundamentals in your chosen one, like Python's Timsort behind the sort call or how Java hashmaps resolve collisions. Fluency in one beats tourism across three.

Comp for context: entry-level Goldman engineers report roughly $111,000 in total first-year pay on self-reported trackers.Compensation varies meaningfully by location and role; for example, the reported NYC average is roughly $133,000.

Say this, don't say that

When you're stuck mid-problem:

Don't say: nothing, while typing and deleting the same line for three minutes.

Say: "My hashmap approach breaks on duplicates. Give me a second to check whether sorting first fixes the key collision."

When asked why Goldman engineering:

Don't say: "Goldman is a prestigious firm with great technology."

Say: "I read about the risk platform work in the engineering blog, and a second-year analyst I spoke with in June described rebuilding a pricing service used by three trading desks. Shipping code that moves real money is the draw."

What about the questions not on this list?

Rotation is constant, so cover the categories these eight represent rather than memorizing them: arrays and two pointers, strings, hashmaps, stacks and queues, light trees and graphs, and one design question like LRU. Recent online assessments have added machine-learning and prompt-engineering multiple choice for some cohorts, worth a skim if your OA invite mentions it. Dynamic programming appears rarely and stays light when it does.

Applications for the 2027 Summer Analyst class open August 15, 2026, and Goldman screens on a rolling basis.. Submit inside the first two weeks, book your OA while the material is fresh, and treat every CodePad mock as a conversation rehearsal. The code gets you considered. The narration gets you hired.