Meta Frontend Developer Interview Questions

“Our photo gallery page felt fine on office machines, but support tickets said it was slow on phones. I used the browser's throttling settings and a low-end test phone, and recorded a performance profile. Most of the time went to decoding large images and running a big script before first paint. I served smaller images based on screen size, lazy loaded everything below the first screen, and split the script so the gallery code loaded first. The result was your result. We then added a size budget check to the build so the page could not quietly grow again.”

“I built an avatar component that showed a profile photo with an online dot. Three teams wanted it. My first version had many props for every case I could imagine, and people still used it wrongly. After watching two engineers try it, I cut the props down to size, photo and status, and wrote short examples. One team needed a group avatar, so I made that a separate component instead of adding more options. When I had to rename a prop, I kept the old name working with a warning for a while. The result was your result.”

“If every post stays in the page, memory grows and each layout pass gets slower. I would render only the posts near the visible area and replace the rest with spacers of the right height. Photos load just before they scroll into view, and each has its size set in advance so the page does not jump. Video plays only when mostly on screen and is paused and released when it leaves. I would measure heights as posts render, since text length varies. When the person opens a post and comes back, the feed restores the same position from saved state.”

“On tap, I update the local state so the heart fills and the count goes up by one, and I send the request in the background. I keep the earlier value. If the request fails, I restore it and show a small message that the like did not save. Quick taps are the tricky part. I would track the latest intended state and only send the final one after a short pause, or tag each request so an old response cannot overwrite a newer tap. When the server answers, I trust its count, since other people may have liked the post too.”

“I would open the React profiler and type a few letters to see what renders. Usually the draft text is stored in a parent that also renders the list of comments, so each keystroke re-renders the whole thread. The cleanest fix is to move the draft state into the reply box component, so only that box updates. If the list still re-renders because of new function or object props, I would make those stable with useCallback or useMemo, and wrap the comment item in memo. Then I profile again to confirm the list is quiet while typing.”

“I would use a web socket for live messages, with reconnect that waits a little longer after each failure. Each message typed gets a temporary id and a status of sending, so it appears at once. It goes into a local queue, and when the server confirms, the temporary id is replaced by the real one. If the connection drops, the queue waits and retries in order. After reconnecting, the client asks for anything newer than the last message it has, to fill gaps. Typing signals are sent at most every few seconds and expire on their own if no update follows.”

“I would treat it as a bug that blocks people, not as polish. First I would turn on a screen reader and use only the keyboard to see what happens. My guess is the new viewer uses plain clickable elements with no role or label, and focus stays on the page behind. I would make the controls real buttons with clear labels, move focus into the viewer when it opens, keep it inside, and return it on close. The escape key should close it. After testing with two screen readers, I would add an automated accessibility check and a keyboard test to the release checklist.”

“I like frontend work where the details decide whether something feels good, like a feed that scrolls without a stutter or a like that responds the instant you tap. Apps like Facebook and Instagram have to do that on old phones and weak networks, which is a harder and more interesting target than a fast laptop. I have built most of my work in React, which came from Meta, and I would like to work close to the people who think deeply about rendering. I want to learn how large teams keep an interface fast while many people change it every day.”