Finished the LinkedIn frontend SWE loop about six weeks ago for a role on the feed infrastructure team. Four YOE, React-heavy background. Sharing because the frontend-specific info I found before interviewing was pretty thin.
The loop was five rounds plus a phone screen.
Phone screen: One frontend-specific problem. They gave me a UI component to build: a paginated list with a search input and debounced API calls. They wanted working code, not pseudocode. I used vanilla JS with a debounce utility because they said framework choice was mine and I didn't want to spend time wiring up React boilerplate. That seemed fine.
Onsite:
Round 1, frontend coding: build a rate-limited button component that disables itself for N seconds after being clicked, handles edge cases (rapid clicks, unmounting). This was in React. They care about cleanup: did you clear your setTimeout on unmount? Do you handle the case where state updates after unmount? Those are the "gotcha" moments they're watching for.
Round 2, algorithms: a medium LeetCode-style problem, nothing frontend-specific. Binary search variant. A bit jarring honestly. Just grind your algorithms, they test them even for frontend roles.
Round 3, system design (frontend): design a collaborative document editor feature (think a lightweight Google Docs inside LinkedIn). This is where it got interesting. They wanted to talk about conflict resolution, websocket connection management, optimistic updates, offline state. Operational transforms vs CRDTs came up. I don't think you need to implement OT, but you should know what the problem is.
Round 4, behavioral: pretty standard, STAR format. They asked about a time I had to push back on a design spec because of technical constraints, and a time I had to ship something I wasn't fully proud of and how I handled it afterward.
Round 5, hiring manager: conversational. Why LinkedIn, what products do you care about, what's your engineering philosophy.
Overall the bar felt like a mid-tier FAANG. Not Meta/Google tier in terms of difficulty, but well above a Series B startup. I'd say prep accordingly: solid algorithms, real JavaScript depth (not just framework surface), and one solid frontend system design.
Got an offer at E4. Turned it down (comp wasn't competitive with another offer I had).