2026 Ramp AI Coding Interview, Experienced Software Engineer Onsite: 3 Experiences
- Company
- Ramp
- Reports
- 3 interview reports
- Level
- Full-time
- Stages
- Onsite
All three of these Ramp onsites had an AI round inside a codebase Ramp provided, and all three candidates were rejected. In one, the repo's Claude.md told the AI to write every prompt to a file the interviewer read afterward.
Coverage: Experienced full-time software engineers: two changing jobs and one recently laid off · The AI coding round at the onsite, plus the tech screen and the other onsite rounds where one candidate described them · Interviews in April–June 2026 (2) and July–September 2026 (1), reported July 9–21, 2026 · No locations given
What came up
In all three reports, Ramp's AI coding round meant building on an existing repo or codebase rather than solving a puzzle from scratch. Two of the three tasks were about matching, and the one candidate who got feedback was told they failed the AI round for not showing enough understanding of what the AI wrote.
| Topic | Reports | Most recent report |
|---|---|---|
| AI round inside a repo or codebase Ramp provided | 3 of 3 | Jul 21, 2026 (interview Jul–Sep 2026) |
| A matching task | 2 of 3 | Jul 21, 2026 (interview Jul–Sep 2026) |
| AI tools set up by the candidate on their own machine | 2 of 3 | Jul 21, 2026 (interview Jul–Sep 2026) |
| Optional bonus items once the main task is done | 1 of 3 | Jul 21, 2026 (interview Jul–Sep 2026) |
| Frontend work on an upload flow backed by an OCR service | 1 of 3 | Jul 18, 2026 (interview Apr–Jun 2026) |
| A finance feature in a larger codebase, short on time | 1 of 3 | Jul 9, 2026 (interview Apr–Jun 2026) |
| Interviewer reviewing the candidate's prompts afterward | 1 of 3 | Jul 9, 2026 (interview Apr–Jun 2026) |
A report can count toward more than one row. Report 01 didn't say what its matching system matched.
Questions we found
AI Coding
| Question | Frequency | All last seen dates |
|---|---|---|
| Build a matching system in a provided repo, any language, then pick bonus items if time is left | 1 report | Jul–Sep 2026 (report date Jul 21, 2026) |
| Redesign receipt upload so a new OCR service extracts the details and suggests the matching transaction, with the user confirming or changing the match | 1 report | Apr–Jun 2026 (report date Jul 18, 2026) |
| Build a finance feature in a larger downloaded repo with local AI, with every prompt logged to history.txt | 1 report | Apr–Jun 2026 (report date Jul 9, 2026) |
Algorithms / Data Structures
| Question | Frequency | All last seen dates |
|---|---|---|
| Retryable URL Maze: explore a maze of linked URLs, with follow-ups on 4xx and then 5xx responses | 1 report | Reports: Apr–Jun 2026 (report date Jul 9, 2026); Additional FastPrep history: Oct 6, 2026; Sep 23, 2026 |
System Design
| Question | Frequency | All last seen dates |
|---|---|---|
| Present a system you designed before, with slides and diagrams | 1 report | Apr–Jun 2026 (report date Jul 9, 2026) |
3 detailed candidate timelines, loops, and screens
01: A matching system, paid for with a $50 tools stipend
Report date: Jul 21, 2026 · Interview period: Jul–Sep 2026 · Role: Software engineer · Experience: changing jobs, came through a headhunter · Result: Rejected
Onsite: AI live coding
- Ahead of the round, Ramp gave a $50 stipend to set up whichever AI tools the candidate wanted.
- Ramp supplied a repo. The job was to build a matching system in it, in any language.
- If the main part got done with time to spare, the candidate could choose a few bonus items to take on.
That's all the author shared about the task. They didn't say what the system matched, what the bonus items were, or how the other onsite rounds went. Their own read afterward: a round like this probably looks past whether you can build the thing, at how you break down the requirements and how you check what the AI generates. They rated the difficulty average.
02: Receipt upload and auto-matching, on a codebase Ramp shared
Report date: Jul 18, 2026 · Interview period: Apr–Jun 2026 · Role: Software engineer · Experience: changing jobs · Result: Rejected
Onsite: AI-enabled interview
- Ramp shared an existing codebase and asked the candidate to improve the UI and UX of one flow.
- The problem with the current flow: attaching receipts is tedious. A user has to find the transaction, open it, and upload receipts one at a time.
- The backend team had already shipped an OCR service. You give it a receipt image, it pulls out the amount, merchant and date, and it suggests which transaction the receipt belongs to. Each call takes a few seconds.
- The task was a better upload-and-match experience: let users upload receipts, show processing status while the OCR runs, display what it extracted, and let the user confirm the suggested transaction or pick a different one.
The candidate called the difficulty average and the overall experience negative.
03: A finance feature against the clock, with every prompt on record
Report date: Jul 9, 2026 · Interview period: Apr–Jun 2026 · Role: Software engineer · Experience: laid off about two months before posting · Result: Rejected
Recruiter call
- About 20 minutes.
Tech screen
- A maze you explore by following links, which the author described as a known question. Practice version: Retryable URL Maze.
- The follow-ups were about failed requests: first handling 4xx responses, then 5xx. The practice version covers retrying 503s and malformed responses, but not the 4xx follow-up.
- After the screen, the process moved over to a different team.
Onsite
Four rounds, which could be done in person or remotely. The author was under an NDA, so they shared impressions rather than exact prompts.
Round 1: Coding on CoderPad
- A practical problem about financial transactions, not an algorithm puzzle.
- The interviewer cared about object-oriented design, clean code and corner cases. The concrete task wasn't described.
Round 2: System design
- The candidate presented a system they had designed in the past, with slides and diagrams.
- It was judged against the level they were interviewing for.
Round 3: AI coding
- The interviewer gave them a repo to download, and they used AI running on their own machine to build a feature in it. The feature was finance-related, much like the CoderPad round, but the codebase was a lot bigger.
- Time was very tight. Waiting for the AI to answer, reading what came back and iterating all ate into it, and there was no time to read every function. The candidate had to keep doing other work while the AI ran.
- The repo's Claude.md told the AI to record every prompt from the session in a file called history.txt. Afterward, the interviewer went through that file to judge how the candidate had worked with the AI.
- The candidate used five prompts in total.
- This was the round they failed. The feedback said Ramp wanted to see more understanding of the AI's output, rather than taking it as it came.
Round 4: Behavioral
- Routine, by the author's account. No questions were named.
They rated the loop hard overall.
Patterns across the reports
-
The work starts from Ramp's code. All three candidates (01, 02 and 03) built inside a repo or codebase Ramp handed them. Two of them brought their own AI setup: 01 got a stipend to pick tools, and 03 ran AI locally on a downloaded repo. Report 02 doesn't say which AI tools were available.
-
Matching and money. Reports 01 and 02 both centered on matching: a matching system in 01, and receipts matched to transactions in 02. Since 01 never said what was being matched, there's no way to tell whether it was the same exercise. Reports 02 and 03 both had finance tasks, receipts in one and an unnamed finance feature in the other.
-
Understanding the AI's code came up twice, in different ways. In 03 it was the stated reason for failing the AI round, and the interviewer had a full log of the candidate's prompts to look at. In 01 it's the candidate's own guess about what the round measures, with no feedback behind it.
Keep practicing
- Retryable URL Maze: the maze from report 03's tech screen, with bounded retries on failed requests.
- Single-Screen Tic Tac Toe: Ramp's one project-coding exercise on FastPrep. It comes from a 2025 phone screen, not from any of the AI rounds above.
- Ramp project coding questions: Ramp's codebase-style practice projects. Sign-in required.
- Ramp system design questions: every Ramp design question on FastPrep. Sign-in required.
- Ramp interview guide: Ramp's interview process and question list.
- Ramp hiring insights: hiring data for Ramp.
- Experience feed: candidate reports from every company.
AI coding rounds at other companies
- Meta: Meta AI coding interview: two onsite rounds where you finish a hard algorithm inside a mostly written codebase with an AI model, then face stress-test suites.
- Google: Google AI coding interview: an online assessment with HackerRank's built-in assistant, and a live round on a multi-file project where using AI was optional.
- Stripe: Stripe AI coding interview: experienced engineers kept getting the same transaction rule parser, built step by step with a built-in model.
- Stripe interns: Stripe intern AI coding round: ten intern accounts of the same AI-assisted round.
- Rippling: Rippling AI coding interview: HackerRank's built-in AI on Rippling's familiar problems, plus an HR codebase at the onsite.
- DoorDash: DoorDash AI coding interview: AI CodeCraft: build and run several small services live with your own AI tool.
- LinkedIn: LinkedIn AI coding interview: one long, multi-part task in an IDE whose agent edits files directly.
- Shopify: Shopify AI coding interview: pair programming with your own AI on Shopify's well-known question set.
- Datadog: Datadog AI coding interview: bring any AI tool and share your screen; the tasks are easy, and the follow-ups decide the round.
- Capital One: Capital One AI coding interview: new grad Power Day agentic round: fix a bug, then add a feature to a large full-stack app with Claude Code.
- Bloomberg: Bloomberg AI coding interview: intern rounds that start in a small file and move to a large codebase, debugging with AI help.
- xAI: xAI AI coding interview: new grad take-homes and onsites where you build with AI.
- Adobe: Adobe AI coding interview: AI showed up as one step in a director round and in a take-home; the rest stayed classic.
- Snowflake: Snowflake AI coding interview: design and build a rate limiter or a retrying task queue while the AI writes the code; one AI Engineer onsite was in person with your own laptop.
- Snap: Snap AI coding interview: a backend-only chat service from an empty Python project with Claude or Codex, or a feature in an existing codebase.
- Anthropic: Anthropic AI coding interview: review four pull requests, then improve one with Claude Code.
- Apple: Apple AI coding interview: CoderPad with a sidebar agent, alongside Apple's other recent interviews.
More companies with AI rounds on FastPrep:
- Amazon: Amazon interview report: an AI-assisted debugging OA (1 experience).
- Atlassian: Atlassian interview report: project coding and AI tool use (11 experiences, two parts).
- Circle: Circle interview report: three AI rounds: building, debugging and PR review (1 experience).
- Volta: Volta interview report: a deliberately vague data task with a token budget (1 experience).
Across companies: 2026 AI Coding OAs at 15 Companies: 32 Reported Problems and 2025–2026 AI Coding Phone Screens and Onsites at 43 Companies: 76 Reported Problems.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.