Rippling Software Engineer Interview Experience — Phone Screen Plus Three Onsite Rounds, Two of Them AI Coding

Overall process: 1 phone screen + 3 onsite rounds. This is pretty much the complete set.

Phone Screen round

AI coding round: the Expense Rules Engine problem that's already on the forum, but with some special follow-ups.

1. What if we have million of expenses and ten thousand rule. How to optimize?
   I answered with batch + aggregate, but you need to discuss basic rules and group rules separately.
2. How do we store the result and notify customer
3. How to optimize, if compound rule is A and B while B is A and C.

3 working days after the phone screen I got the message that I was moving on to the onsite.

Onsite

Round 1: System Design: Design metrics aggregation.
There are lots of events (like user clicks) and many event types, and you need to design a system to take in the events, and then a dashboard that queries aggregate counts. I'd suggest getting familiar with this ahead of time. The fourth requirement never got discussed because we ran out of time.
Requirements
Support event data collection (1 M events and 10K event type)
Support dashboard query like click event count within a day
Support context adding to event data like using ip from client to identify location of event
Support complex query like click event count in US region in last 3 months

Round 2: AI coding round (Web API)
You're given an HR system that supports employee info. The database stores employee data as well as employee historical data, and the feature to add is being able to schedule an employee's change, for example a promotion next month, so you need to schedule a title change. Both frontend and backend need changes, but you can focus on the backend.
This round is probably where I failed. At first I was thinking of adding a table, with a backend worker pulling the changes that are about to come due and applying them. But the interviewer didn't think that would work. So I switched to storing the history of changes and applying the changes that have reached their time when reading. But you have to be careful that a new change doesn't overwrite the fields of an older change. Anyway, discussing the solution took so long that I ran out of time to fix the bugs.

Round 3: Project Dive Deep
I prepared a 3-page PPT. It was pretty deep, and they asked a lot about trade-offs. There was only one extra behavioral question: Tell me about a time when you provide critical feedback to another people