Went through the LinkedIn L5 (Senior SWE) loop earlier this year and the system design round is where I felt the most pressure. Putting down everything I remember.
They gave me a prompt right at the start, no small talk: "Design LinkedIn's feed ranking system." 45 minutes. One interviewer. They shared a virtual whiteboard.
What they actually cared about: Clarifying the problem before jumping in. I spent the first 4-5 minutes asking questions (DAU scale, read vs. write heavy, latency requirements). They were visibly pleased. Your data model. They asked me to walk through the schema for posts, users, and rankings explicitly. Where you'd put caching and why. Not just "Redis" but what goes in Redis and what the eviction policy would be. Handling the "hot post" problem. If something goes viral, how does the system degrade gracefully.
What they didn't push on hard at L5: I didn't have to go super deep on consensus algorithms or distributed transactions. The vibe was more "can you reason about tradeoffs" than "prove you've memorized Raft."
I was also asked a follow-up about how I'd A/B test the ranking model, which caught me slightly off guard but makes sense for LinkedIn specifically given how important feed relevance is to their business.
At the end they asked how I'd monitor this system in production. SLOs, alerting thresholds, what a degraded state looks like. That part was 8 minutes or so.
For senior system design at LinkedIn in 2026, I'd say: own the scope, be explicit about your assumptions, and always tie your choices back to user experience or business outcome. Generic "scale to millions of users" talk without LinkedIn-specific context doesn't land as well.