I interviewed with Rippling's new AI department, which has 3 teams and an already-launched product. Their main product is an internal chat agent system that can integrate with hundreds of integrations on top of employee data to answer all kinds of questions.
Recruiter - 30 minutes. Talked about which project of mine stood out to them, my involvement in it, and whether I had led any projects of my own.
System design - 60 minutes. The question was to design a counter like a Prometheus/Datadog counter. Clients send different kinds of events to the backend through their own SDK. Clients include mobile and web. You need to account for a phone possibly sending a big batch of events all at once because it was offline. No event can be lost. The goal is to support rolling these scattered events up onto a dashboard (with a data-freshness requirement of a few dozen minutes), and also to compute billing and total event counts every month. Each event also needs tags including timestamp, department, project, and other attributes. Events can be duplicated and delayed (phone goes offline, then sends everything at once when it comes back online).
HM - 60 minutes. 40 minutes was a project deep dive, 10 minutes was leadership & behavior. You're allowed to use a 3-slide deck, and you can draw your project/system diagrams ahead of time on Excalidraw to save time. Most of the questions were about how you'd scale things — if traffic and volume went up 100x, what would you change — and whether you'd design it differently if you did it again. My project involved an AI agent, so they asked a lot about handling real-time hallucination, tool-use permissions, token usage, and switching between different language models. The behavioral part was the usual stuff: how you lead a team, how you give feedback, what kind of environment you like and dislike.
Onsite
Coding - 60 minutes, 2 levels. The question was to calculate a driver's work hours and pay: 4 APIs to write — add a driver and their hourly rate, record a driver completing a trip, and calculate total payout. You had to finish it, write tests, discuss the data structure tradeoffs, and model it with OOP as much as possible. There were quite a few edge cases you needed to clarify, like time ranges that already overlap, how a change to the hourly rate affects things before and after the change, and the time/money formats. Level 2 was to calculate which wages had already been paid and when, and which hadn't. The interviewer never told me beforehand what level 2 would be, so once that question landed I realized I had to go back and rework my own class model.
System design again - 60 minutes. Design an employee-termination workflow: one endpoint that, behind the scenes, calls several different systems and services — canceling the employee's credit card, deactivating accounts, scheduling final payment, revoking access (email, Slack), transferring data, wiping phone data — roughly 7-10 of them. Some systems can't be safely retried (a credit card can't be canceled twice), some steps have to happen first (like deactivating the account), some can be done in parallel, some have priority ordering, some systems support the saga pattern, some have webhooks. It's mostly about correctness and completeness rather than heavy traffic, but you also need to minimize completion time, and if the process stops partway through, it should be able to automatically resume from wherever it left off.
Practical coding/build - 90 minutes. They had me build a simple LLM chatbot that could answer questions about company employee expenses. The frontend lets you ask questions in plain language, and the backend turns the question into database parameters — you're allowed to use AI for this, but the interviewer can see the whole process of how you use it, and I think that was part of what got scored. The frontend was mostly already written. There were 2 big JSON files provided as the "database" — one for companies and employees, one for employee expense records. Level 1 was handling a few basic questions and designing the API; the JSON had employees and departments across different companies. Level 2 was doing aggregation — which company or employee reported the most expenses, which department, at what time. There was a level 3 but I never got to see it, time was very tight. Along the way you also had to discuss your direction and strategy with the interviewer: whether to do several LLM passes or build an agent. You also had to implement streaming. The prompts and instructions had to be discussed and then written by you.
Company values + behavior - the recruiter sends prep material beforehand; basically you need to have your own projects and experiences ready that demonstrate and prove out Rippling's core values. I was asked what situations or experiences changed my mind or changed the direction of my project, how I know when I'm wrong versus right, whether I've had experience handling things outside my own responsibilities, and to share the most complex production issue I dealt with and what I did about it.
Discussion
Loading comments…