Just finished a Datadog frontend SWE loop and want to share the shape of it because I went in expecting standard React/JS and got more infrastructure-adjacent questions than I expected. Not bad, just different.
I'm a 4 YOE frontend eng, primarily React + TypeScript, Bay Area. Applied for a senior frontend role on one of their product visualization teams.
The loop was 4 rounds after an initial technical screen.
Technical screen: 45 min. One JS/TS coding problem (medium difficulty, had to implement a custom hook with some tricky async behavior) and a conversation about state management patterns. I'd say know hooks deeply, not just useState/useEffect. They probed on useReducer and how you'd handle complex derived state.
Round 1 on-site: Frontend systems. They asked me to walk through how I'd build a real-time metrics dashboard. Not just "what components would you use" but how do you handle the websocket connection, how do you manage update frequency vs DOM thrash, how do you paginate time-series data in the UI. This is a very Datadog-flavored question and I should have prepped for it more. Think through performance constraints specifically.
Round 2: Coding. Implement a debounced search input that handles out-of-order API responses. Classic but they wanted the full implementation: TypeScript, error states, loading states, cleanup on unmount. I got through it cleanly.
Round 3: Accessibility and design system. Surprised me. They asked about ARIA, keyboard navigation patterns, and how I approach component library decisions. If you've never thought about a11y seriously, spend a week on it.
Round 4: Behavioral. Standard. Focus on times you pushed back on a design decision, navigated ambiguity, or handled a prod incident related to frontend.
The vibe throughout was that they want frontend engineers who care about correctness and performance, not just shipping UI quickly. Made sense given their product is basically "display massive amounts of data without melting the browser."