I interviewed for a Meta DSA onsite a while back and got the offer, so I want to write this down while I still remember it.
My takeaway from the interview is that the real thing differs from interview reports in the details, so when you prepare you still need to keep your thinking flexible. Grasp the big picture, but don't just memorize answers. The interviewer wants a fairly quick response — basically as soon as they give you the question you need to start organizing your thoughts and talking, so regular practice matters a lot.
AR:
A social network company wants to roll out a new ads ranking algorithm believed to be better than the current one.
- How to evaluate the new algorithm
- How to run experiments
- How to interpret results
- How to balance tradeoffs between revenue, advertisers, and user engagement
- How to communicate recommendations to leadership
AE:
Chatbot LLM model evaluation — similar to a case I'd come across before. Even though it's called AE, most of the time was actually spent discussing the case, not doing calculations right away.
Chat Bot
Two LLM models.
- Define a good response from a chatbot as being both helpful and honest. P(helpful) = 0.8 and P(honest) = 0.9, what is P(good)?
- What is the probability of getting two good responses in a row?
- Given the first three responses are good, what is the probability that the 4th one is also good?
SQL:
Active Ads — similar to a case I'd come across before.
BQ:
Basically all the questions had been posted before, so just prepare 6 or more stories/examples. The interviewer cared a lot about impact.
I got notified that I passed two weeks after the interview ended. Good luck everyone!
Discussion
Loading comments…