- Frequently Asked Questions
- How Coinbase scores Software Engineer candidates
- Recruiter screen
- Structured assessments
- Technical screen
- Coinbase Software Engineer Onsite Interviews
- 1. Tech Execution
- 2. Domain Deep Dive
- 3. System Design
- 4. Reverse System Design
- 5. Foundational interview
- Offer review and executive approval
- Coinbase Software Engineer Total Compensation (USD)
The Ultimate Coinbase Software Engineer Interview Guide
This interview guide covers the loop for Software Engineer candidates interviewing at Coinbase. It's been written with the help of 4 Senior Software Engineers and 2 EMs who have worked at Coinbase, who are coaches on Prepfully. Thanks to candidates who are interviewing there in real-time, we've managed to keep this up-to-date with questions asked as recently as last week. This guide covers the full loop and lists recently reported questions for each round, including the AI Fluency rating Coinbase has attached to every engineering candidate since 2026.
The Coinbase Software Engineer interview loop typically goes through these stages:
- Recruiter screen
- Structured assessments:
Cognitive and culture assessment
CodeSignal coding assessment - Technical screen (some loops)
- Onsite, up to 4 virtual 1:1 interviews drawn from:
Technical Execution
Domain Deep Dive
System Design (IC5+)
Reverse System Design (senior and Staff roles)
Foundational, with HM - Offer review and executive approval
Before you even meet this long list, your resume and LinkedIn profile are read against the role and Coinbase says only about 5% of applications make it past this review.
How Coinbase scores Software Engineer candidates
Every interviewer scores you against Coinbase's cultural tenets. So we’d recommend ensuring you tick as many of these across your answers:
- Clear communication: being direct and succinct while sharing what people need before they have to ask, and listening actively
- Efficient execution: have a bias for action while striving for quality and focusing on the 20% of work that gets 80% of the impact
- Act like an owner: own outcomes even when you worked on the project only partly
- Continuous learning: are humble and care more about getting it right than being right and treat every setback as a chance to learn
- Top talent: whether you’d raise the team average both in skill and culture
- Championship team: whether you hold yourself to high standards and push outside your comfort zone to “be a part of a winning team that isn't a family”
- Customer focus: whether you solve customer problems with solutions that are easy to use and can be trusted
- Repeatable innovation: if building and innovation is a habit of yours where you prioritize shipping solutions instead of just discussing them
- Positive energy: optimism about the future and the determination to get there, solve problems together without blaming anyone
- Mission first: keeping the workplace focused on the mission, with no social or political activism at work that's unrelated to it
Since 2026, every stage of the loop also scores candidates explicitly for AI fluency. This is across 3 dimensions:
- Usage: using AI tools effectively to get measurably better outcomes
- Application: knowing when AI is the right solution and when it isn't
- Understanding limits: using human judgement as a guardrail and knowing where AI breaks down or can be dangerous, especially the privacy and security implications
Recruiter screen
On this call, you’ll either meet a human recruiter or Coinbase's AI Recruiter. You’ll discuss your (crypto) experience and alignment with their mission so that your level and compensation can be settled.
A Coinbase candidate who took the AI Recruiter screen reporting that they ran out of time before they had covered every question, Blind
Expect questions like:
- Tell me about the biggest project you’ve worked on.
- What’s your experience with crypto so far? Where does your interest come from?
- Which of our cultural tenets show up the most in the way you work?
Use this call to sort out the details of the rest of your loop and how much AI you can use in each round because these details change from loop to loop.
Structured assessments
These are online tests you take before the onsite loop that are made up of a cognitive and culture assessment alongside a CodeSignal coding assessment.
Candidates describe the cognitive section as 50 multiple-choice questions in 15 minutes with no penalty for wrong answers. The culture section is a workstyle questionnaire about how you carry yourself at work. Candidates' accounts describe this portion to be a marathon of over 100 questions where a couple of them come back reworded.
The cognitive and culture assessment measures
- logical and spatial reasoning
- verbal reasoning
- mathematical reasoning
- how well you’d thrive in Coinbase's culture
As an engineering candidate, you’ll also complete a CodeSignal coding assessment as proof of your technical abilities. It’s a single problem that has 4 progressive levels. Each level extends the code from the previous one with new requirements, and the levels unlock sequentially. Since 2026, the standard coding assessment has been replaced by an AI-Assisted Coding Assessment where you use AI tools to solve problems. For instance, a recent Prepfully candidate reported working with the AI-assistant to write the jump logic for a Flappy Bird game that the platform drew on a simulated map.
Technical screen
This is a 60-minute live coding interview with a Coinbase engineer, and is the gatekeeper for most candidates to the onsite. Some loops skip this and send you straight to the onsite.
You’ll be briefed about a practical problem, like packing a block from a mempool, and then you’ll have to implement it in the language of your choice.
Once again, the problem is progressive so the interviewer adds requirements as you go. To continue the mempool example, part 1 asks you to fill a size-capped block with the transactions that earn the most fees. Part 2 adds parent-child dependencies, so a child transaction can only enter the block alongside its full chain of parents, with each parent placed before its child.
The interviewer will assess:
- The end result and if it’s production-grade code
- How you approached the problem and refined it
- How you respond and integrate the interviewer’s follow-ups and feedback
Coinbase says you don't need crypto knowledge unless your role calls for it, and your recruiter can tell you whether yours does. Candidates say you need to be comfortable with crypto fundamentals (how mining and CPFP work) because prompts can be hard to follow without a basic understanding of these terms.
Coinbase Software Engineer Onsite Interviews
During the onsite, you’ll face:
- Tech Execution
- Domain Deep Dive
- System Design (IC5+)
- Reverse System Design (reported at IC6+)
- Foundational interview
1. Tech Execution
This is a 60 or 90-minute live coding round conducted in the same progressive format as the technical screen. The problems are usually medium difficulty and are presented as a practical scenario like an order system or a leaderboard. You’ll start off with an initial build and then face 2 followups.
In some loops, you’ll get started in a blank file with no starter code where you’ll have to write your own tests too. A couple of candidates report that the round ran in Coinbase's own online environment where they could run tests and look stuff up in the docs.
Starting March 2026, you’re allowed to use documentation and AI tools during the round but the extent to which it's permitted isn’t standardized yet so please confirm with your recruiter. A Staff candidate said they were allowed to use AI to clarify the problem but had to type every line of code by hand.
Prepfully candidates who appeared for this round reported the following questions:
- Build an order management system where each order can be live, paused or cancelled
- Design a leaderboard that supports adding scores, removing users and returning the kth-highest score
- Implement a moving average over the last k values
- Write an iterator that round-robins across multiple lists
- Build a matching engine for buy and sell orders
I suggest that you treat this hour as a demo of your engineering technique. Start simple and build from there so it gives you an opportunity to keep checking in with the interviewer and be sure that you’re headed the right way. If you get called out, or if something doesn’t work - always acknowledge it by verbalizing and talk through how you plan to fix it. How you approach your mistakes and recover from them is a really important competency signal in its own right.
2. Domain Deep Dive
This is a 60-minute technical interview and the format varies a lot by loop. Some loops run it as a second coding round and some as an AI-assisted round where you’ll work inside an existing repo.
If run as a second coding round, the format matches Tech Execution. In the AI-assisted version, the tasks are pretty much what you’d tackle day-to-day in a codebase.
Coinbase uses this format to understand how you’d
- triage and debug failures in code you didn't write
- review code for correctness and performance
- think through backward compatibility and rollback and
- validate AI-generated output if you’re using it in the interview
Expect questions like:
- In this codebase built around database cursor queries, find the bug, talk through how you'd build a new feature, review this pull request, and debug this production issue, using the AI assistant.
- At what point would you stop trusting the assistant’s suggestions for this codebase?
- The author says this change is a pure refactor. Is it?
If you’re working with an AI-assistant, be extra vocal about your thought process and narrate every decision as you go. How you direct and critique the AI makes up for so much of the signal the interviewer is gathering and they can’t credit you for your reasoning unless they hear it.
Kyle Cesmat, Engineering Manager for Agent Experience at Coinbase, says one of his core tenets for AI in development is that “we are essentially asking these systems to produce code on developers’ behalf. So developers are still taking responsibility for the code that is produced... AI should never be making decisions.”
3. System Design
This is a 60-minute design conversation where you’ll design the architecture of a real-world service and is typically applicable to IC5+ candidates. Expect the discussion to be rigorous about the trade-offs you chose and the corresponding effect that they’d have on the business.
Interviewers will be grading your performance on:
- how you scope the problem and set requirements
- how well you articulate and explain your thinking behind the design
- whether your design would survive production
The prompts are built around Coinbase's systems itself so you’ll benefit from practicing on problems where money moves and correctness under concurrency matters more than raw scale. Expect prompts like:
- Design a trading system that routes orders to a third-party exchange
- Design a price alert service
- Design an order book matching engine
- Design a real-time price aggregation service across exchanges
Generally speaking, it’s super important to get the basics right in this round. After asking targeted questions to set the requirements for the design, describe the system at a high level before you draw anything. As you go into the details, address the tradeoffs you’re making at each point proactively before the interviewer has to hint about it.
Only go deep on a component if the interviewer steers you toward it. If they add a constraint or change a requirement, adjust the design and don’t get defensive. If you don't know something about a component, say so honestly but talk about how you'd try to find out more. Being confidently wrong will definitely work against you.
For each piece, pick a technology you've used and say why it makes for an apt choice. Talk through how the design scales to what Coinbase needs and how it works in prod. Before you wrap up, double-check if you’ve answered the problem completely.
The System Design format is somewhat standardized across the industry, so we will not go in-depth within this guide; please refer to our System Design guide for more details.
Design sessions are almost always the biggest driver of the seniority / levelling decisions for engineering roles, so we recommend that you sit down for a 1:1 session with our coaches so that Coinbase’s own SWE interviewers can simulate this for you. They’ll conduct the round just like they do at Coinbase and then give you targeted feedback and even prep plans so you can have some hefty leverage on your end.
Fill out a quick form to get matched with an expert or book directly with a Coinbase SWE interview coach of your preference.
4. Reverse System Design
In this round, you’ll be given a technical design doc from a sister team that clearly has some gaps and issues and by the end of the hour, it’s your job to work through trade-offs and patch up the doc into better shape.
The round runs in “reverse” because you’re already given the solution and you have to work backwards to the problem. You’re expected to lead the conversation but the interviewer will frequently steer you. For instance, a Prepfully candidate got asked “What's not idempotent in this design but needs to be?”
Some loops run this as a project retro where you’ll walk the interviewer through a system you’ve already built and defend your decisions when the interviewer probes.
About your own project, expect the interviewer to ask questions like:
- Why did you pick this database?
- If you rebuilt this system today, which parts would you hand to an AI agent?
- Does this system have some kind of dependency on you in a way that only you can fix?
None of what we’re going to say next is news to you but is worth saying anyway because this format rewards discipline over instinct. You may notice some obvious glaring errors while reading the doc but you have to resist the urge to dive right into it. Open with questions about the problem this design was meant to solve to lay some context so that you can keep a running order in your head of high level to low level fixes. Generally speaking, keep an eye out for security, observability and the non-functional requirements because they are the things most likely to be missing from a doc that otherwise looks sound.
5. Foundational interview
This is a behavioral interview with a hiring manager who’ll ask questions about your experience and past work to understand:
- how you make decisions
- how you decide what is worth building
- how you handle disagreement
- how you build trust across a team/org
- how you raise the bar (for yourself and the overall business)
Expect questions like:
- Tell me about a time you shipped an MVP. What did you do to keep improving it?
- How do you judge whether you’ve written enough tests? Are there any instances you think it’s acceptable not to write tests?
- Have you ever disagreed with a team decision/direction? What did you do about it?
- Have you ever strained a working relationship?
- Have you ever taken the lead on a project where nobody had the full picture?
In the behavioral rounds, you need to land on the right story quickly and recollect details without dragging out your story. The best way to do that is to write a story bank of 6-8 stories before the loop and practice their delivery. A set like this covers 80-90% of the questions you’ll get.
We recommend that you take the classic STAR framework and build it out into STAR+ (Situation, Task, Action, Result + the X-Factor: Learning) to structure your stories. You end each story with a personal takeaway and how it changed the way you work. This way, you’ll demonstrate intellectual humility which implicitly links to their cultural tenets, especially Continuous Learning.
Offer review and executive approval
Everyone who interviewed you during the onsite meets to share their feedback on you. The panel doesn’t have to agree unanimously for you to get an offer. If an interviewer does raise a concern, though, the panel has to work through it and address it before your offer can go any further. After the panel has discussed their points, the hiring manager decides whether to move forward with an offer. For some roles, an executive makes this call instead.
If the answer is yes, the HM puts together an offer packet which includes a one-paragraph summary of how you’ve shown extraordinary ability.
Since 2026, every engineering candidate also gets an AI Fluency rating of Above, At or Below Threshold, based on how they did across the loop and the executives can see this rating when they review your offer.
The packet then goes to Coinbase’s CEO or COO who reviews every offer Coinbase makes. Your offer reaches you only after one of them has signed off on it.
Coinbase Software Engineer Total Compensation (USD)
- IC3: $206K
- IC4: $264.7K
- IC5: $398.5K
- IC6: $568.2K
Recently reported Coinbase Software Engineer interview questions
If your search operations are suddenly running in linear time what happened to your tree
Design a Distributed Metrics Logging and Aggregation System.