Rippling Rules and Task Screens: 22 Experiences (Part 3 of 12)
- Company
- Rippling
- Reports
- 22 interview reports
- Level
- Full-time
- Stages
- Phone screen
Rippling phone screens on the expense rules engine, task manager, pluggable logger and more delivery costs, 2025 to 2026.
At a glance
- Asked: The expense rules engine in 4 of 22 reports, the task manager in 4, the logger in 3.
- Format: Entries 37, 32 and 34 were AI coding screens, and entry 37 says the round tests planning and prompting more than hand-coding.
- Where it goes wrong: In entry 32 the interviewer changed direction while the plan was reviewed and the AI code never ran; in entry 34 the AI missed that a filtered parent's children must be hidden too.
Coverage: 22 reports, Jan 2025 to Oct 2026. 10 more delivery cost screens, then 12 screens on other practical tasks. Levels: SDE-2 to Staff, plus reports with no level given. Places named: the U.S., India and the U.K.
Terms: OA = online assessment · HM = hiring manager · LLD / HLD = low-level (class) / high-level (system) design · VO = virtual onsite · DSA = data structures and algorithms · CodePair = HackerRank's shared coding pad.
In this series
- Part 1: overview, question tables by role, what to practise
- Part 2: delivery screens
- Part 3 (current): delivery screens, rules and task screens
- Part 4: practical screens, music and key-value screens, no-task screens
- Part 5: no-task screens, design screens, recruiter and manager rounds
- Part 6: senior loops
- Part 7: senior loops
- Part 8: senior loops, Staff loops, SDE-2 loops
- Part 9: SDE-2 loops, SDE-1 loops
- Part 10: no-level loops, History
- Part 11: History
- Part 12: frontend and mobile, data, History
What came up
These screens are small business systems: rules over expenses, a task list, a logger. Counts over the 22 reports here:
| Topic | Reports | Last seen |
|---|---|---|
| Delivery cost tracker and variants (2 variant or unlinked) | 10 of 22 | Jul 2025 |
| Expense rules engine | 4 of 22 | Oct 2026 |
| Task manager filter and sort | 4 of 22 | Sep 2026 (report date) |
| Pluggable logger | 3 of 22 | Oct 2026 |
Rows overlap: a report can count in more than one row.
Delivery cost phone screens (continued)
22: Four delivery parts in Python classes
Interview date: Jul 2025 · Report date: Aug 12, 2025 · Role: Senior Software Engineer · Result: No offer
- Recruiter.
- Phone screen: Four delivery parts in Python classes, where the third needed a more complex algorithm.
- Driver deliveries in four parts (first two easy, third needs a more complex algorithm).
23: One slice of a delivery app, AI allowed
Interview date: Jun 2025 · Role: Senior Software Engineer · Location: New York · Result: No offer
- Practical coding: one slice of a delivery app, the cost side. AI and web search were both allowed.
24: Staff: delivery costs with O(1) queries
Report date: Jun 13, 2025 · Role: Staff Software Engineer
- Recruiter (45 min): A talk about their work.
- R1 coding: Delivery costs with payments.
- Delivery Cost Tracker: Drivers with float rates and deliveries. GetTotalCost O(1). PayUpToTime pays deliveries completed before time. GetCostToBePaid O(1). How they solved it: Queue of deliveries removed when paid.
25: Delivery costs by hand, part 3 talked through
Report date: May 26, 2025 · Role: Senior Software Engineer · Location: Bengaluru · Result: Rejected
- Recruiter (Zoom).
- Technical screening: AI was allowed if you said so up front, and was judged by different criteria. There were three parts, and they did not use AI.
- Delivery Cost Tracker: AddDriver, addDeliveries(start, end) and a fast getTotalCost. Then payUpToTime and getCostToBePaid (they gave an O(n) version and a faster one was wanted). Part 3, approach only: deliveries happening at the same time across all drivers. How they solved it: Sorting and overlap check for part 3.
26: Told cleared, then rejected
Report date: May 23, 2025 · Role: SDE 2 · Result: Told cleared, later rejected
- Phone screen: Recruiter first said cleared, HM pending, later rejected.
- Delivery Cost Tracker: Delivery management system (design-style coding).
27: Delivery costs extended three times
Interview date: May 2025 · Report date: May 7, 2025 · Role: Senior Software Engineer · Result: No offer
- Live coding.
- Delivery Cost Tracker: Total delivery cost across many drivers with different hourly rates and overlapping paid deliveries. Records arrive immediately. Total shown on accounting dashboards. Extended three times.
28: Grocery delivery methods on HackerRank
Interview date: Apr 2025 · Role: Software Engineer 2 · Location: United Kingdom · Result: No offer
- HackerRank: A class with a few basic methods for grocery delivery.
- Write a class with a few methods for grocery delivery (the driver cost family).
29: Delivery costs in CodePair
Interview date: Apr 2025 · Role: Senior Software Engineer · Location: Seattle · Result: No offer
- CodePair (60 min): Code that had to run, built from step-by-step prompts.
- Delivery Cost Tracker: A delivery service: look up a driver's rate, log a delivery, and total the cost of all deliveries.
30: Food delivery API, stopped at concurrency
Interview date: Mar 2025 · Role: Software Engineer II · Location: India · Result: No offer
- Coding: Told to stop on concurrency.
- Delivery Cost Tracker: Food delivery: addDriver(id, pay_per_hour), recordDelivery(id, start, end), aggregate cost.
31: Delivery APIs and why floats are bad for money
Report date: Jan 16, 2025 · Role: Senior Software Engineer · Result: Rejected
- First DSA round, in CodePair: delivery app endpoints: addDriver(id int, hourlyRate float), addOrder(driver, start, end) with a reason for the time type chosen, and getTotalCost.
- Follow-ups: why float and double are risky for money and what to use instead, then how to let a driver's hourly rate change.
- They finished one of three parts and were rejected.
Rules engine, task manager, logger and other practical screens
32: An expense-rules screen where the target kept moving
Interview date: Oct 2026 · Report date: Oct 26, 2026 · Role: Software Engineer · Result: Rejected
- AI coding screen on HackerRank with its agent switched on and the screen shared. The interviewer changed the day before. Python setup ate five to seven minutes.
- The task was the corporate card expense rules checker. Its last line said a later step must let people add rules through an API.
- Their process: clarify the requirements, sketch the structure in plan mode, review the plan, run it, then review the code. While they reviewed the plan, the interviewer first accepted a rule model built from an operand and an operator, then said a hard-coded check function was fine.
- Time ran out before the scale discussion. The AI-written code had import errors and never ran.
33: A pluggable logger with four handlers
Interview date: Oct 2026 · Report date: Oct 19, 2026 · Role: Software Engineer · Result: Rejected
- Technical phone screen, AI allowed (they used Cursor). A logging tool: one Logger class that runs several handlers, where new handlers can be added without touching the Logger (Stateful Log Transform and Search).
- There were four handlers to start. One removes every copy of a set substring and prints. One cuts the message to N characters and prints. One upper-cases it and prints. One stores it in a list instead of printing. Each handler works on the message it receives, independently. So logging "hello world" prints several versions and also stores one.
- The design had to stretch to future handlers, for example one that opens a database connection per log, or one that sends the message to a remote API.
- Follow-ups: how to optimize, and how to make handlers easy to pass in by hand. They finished but were rejected. They had broken the problem down and written pseudocode first, and found some questions confusing.
34: Task filtering where the AI missed the parent rule
Interview date: Jul–Sep 2026 · Report date: Sep 3, 2026 · Role: Software Engineer · Result: Rejected
- AI coding phone screen, which the candidate rated hard: the task manager filter, sort and dedupe exercise.
- The AI overlooked an edge case: hiding a parent task has to hide its children as well. The candidate's four-task test had an assigned grandparent that is hidden, so its child and grandchild vanish too, while an unrelated task stays visible.
35: The logger again, with a slow model
Interview date: Jul–Aug 2026 · Report date: Aug 31, 2026 · Role: Software Engineer · Result: Rejected
- Technical screen: design a Logger whose handlers are independent of each other (Stateful Log Transform and Search). It took them a while to notice that each handler stands alone.
- AI coding with Claude Opus. The output was slow during the interview, and they suggest picking a faster model.
36: A task scheduler that prints parents first
Interview date: Jul 2026 · Report date: Jul 7, 2026 · Role: Software Engineer · Result: Not stated
- Technical screen with AI allowed. Looking back, they think coding alone would have been faster.
- Part 1: each task has an id, due date, create time, a high-priority flag and an assignee. Filter by assignee, drop finished tasks, then sort by due date, breaking ties by the priority flag and then by create time. Print the list, and when a task has a parent, also print the parent's description (Filter, Sort, and Deduplicate Scheduled Tasks).
- Part 2: add a parent id. Run the parent before its child and print the order the child tasks run in.
37: Expense rules, mostly explained out loud
Interview date: Jul 2026 · Report date: Jul 1, 2026 · Role: Software Engineer · Result: Not stated
- Phone screen: the expense rules check. Most of the hour went to explaining: how to model rules and how the examples flow through. AI wrote the evaluate function at the end.
- The interviewer was quiet and gave little feedback. The candidate's takeaway: this round tests planning and directing the AI more than hand-coding, and using AI is the default.
38: Prepared three questions, got a fourth
Interview date: Apr–Jun 2026 · Report date: Jun 25, 2026 · Role: Software Engineer · Result: Rejected
- Applied online. A friendly recruiter call, then one technical screen. AI was optional but strongly suggested; bring your own environment and AI tool.
- They had prepared the delivery, expense and camel card questions and got none of them. About 5 minutes of intro, 45 of coding and 5 of chat.
- The task was a task manager: part 1 filters and then sorts by the stated rules; part 2 uses a more complex sort order.
39: Task manager rules spelled out, interviewer distracted
Interview date: Apr–Jun 2026 · Report date: Jun 18, 2026 · Role: Software Engineer · Result: Rejected
- AI coding screen, AI could not be turned off. The interviewer seemed busy with their phone and slow to answer questions.
- The task manager took an unsorted list. Two tasks with the same description and due date are duplicates; keep the first. Keep tasks that are unassigned and not done. Order by due date, then the high-priority flag, then oldest create time. Print each task's id and description, and add the parent's description when there is a parent.
40: A rate limiter finished too fast
Interview date: Jan–Mar 2026 · Report date: Jun 10, 2026 · Role: Senior Full Stack Engineer · Result: Rejected the next day
- HR call, the phone screen the next day, a rejection the day after. HR said the screen was AI coding and needed a local IDE and SDK set up.
- After a short chat, the task was a rate limiter. They were told AI was fine and just to get it written, with no word on how it was judged. They used Cursor with Gemini Flash and finished the first version quickly.
- The follow-ups felt made up on the spot. The candidate's lesson: with AI you can finish too fast and leave the interviewer with nothing to judge, so explain why the AI wrote what it did and talk through trade-offs.
41: The logger without AI, then keyword search
Interview date: Apr–Jun 2026 · Report date: Jun 9, 2026 · Role: Software Engineer · Result: Rejected
- Technical screen. Unsure of the AI format, they asked to switch AI off and wrote it alone.
- A logger with handlers, such as a capitalize handler and addHandler.
- Follow-up: search the logs for a keyword. They first walked through scanning every log, then were asked to optimize: index each word when a log is added so search is a direct lookup.
- The interviewer later said the bar is very high.
42: A rule evaluator with test conditions that changed
Report date: May 16, 2026 · Role: Senior Software Engineer · Location: London · Result: Rejected after round 1
- Coding round 1 (AI enabled): Claude was allowed, and they were short on time at the end.
- Corporate Card Expense Rules: Rule evaluator (AI-enabled coding): evaluate rules against inputs; interviewer asked for example input structure and changed test conditions mid-way.
43: Expense validation on HackerRank
Interview date: May 2026 · Report date: May 28, 2026 · Role: Software Engineer II · Result: No offer
- Recruiter.
- HackerRank technical screen: Failed.
- Corporate Card Expense Rules: Validate expenses against supplied rules.
Patterns across the reports
- Explaining and coding. Entry 37 spent the round explaining and had AI write the evaluate function at the end; entry 44 had about 5 minutes of coding after 50 to 55 of discussion.
- Extensible handlers. In entry 33 new handlers had to be added without touching the Logger class.
Common questions
What is the Rippling expense rules question?
Given expenses (id, trip, amount, type) and rules such as a cap per expense or per category, return the violations (46 and 47); entry 32 added rules through an API later.
What is the Rippling logger question?
A Logger with independent handlers (remove a substring, truncate, upper-case, store) in entry 33, then keyword search over stored logs in entry 41.
Keep practicing
- Delivery Cost Tracker: see reports 23, 24 and 25.
- Corporate Card Expense Rules: see reports 32, 37 and 42.
- Filter, Sort, and Deduplicate Scheduled Tasks: see reports 34, 36 and 38.
- Rippling question banks, sign-in needed: coding, behavioral, system design and project coding.
- The Rippling interview guide, Rippling hiring insights and the Experience feed.
- More roundups: intern, new grad, no level and AI coding rounds.
In this series
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.