Meta MLE ML Specialization and ML Design Onsite
Hi. I have an upcoming onsite for MLE at Meta (E4) and wanted to get some information on what type of things to expect in the ML Specialization and ML Design interview, and how best to prepare for them. I dont have a system design interview (instead going for ML Specialization) as I have more experience working in applied ML during my PhD. What I have so far for ML Design: -Grokking Machine Learning at Educative - Videos on youtube on broad ML Design interviews. For ML Specialization I have absolutely no idea on what type of questions, and what sort of indepth discussion is required. Any directions, on both of these interviews, from people who have appeared or who interview people in this role would be really appreciated. thanks #meta #fb #mle #facebook #machinelearning #machinelearningengineer #engineering
If it’s SWE, ML and the specialization isn’t research the ML specialization will most likely be just another ML design round but you’ll have to go more in-depth. Not as in-depth as you’re probably thinking but you’ll talk about that instead of infra and deployment stuff
thanks for your help. yes it is Machine Learning Engineer, so it is SWE. The broad sense my recruiter provided me was that ML design is more about how, and ML specialization is more about why. When I discussed switching the system design to ML specialization he mentioned that the specialization interviews are generally for higher level positions but the option is available in my case. Hence I didnt know how much detail they expect. i.e should I, for example, be able to tell the broad difference between choosing between random forest and gradient boost, or would I be asked to write loss function and do statistical derivations? Do you have any sources/ general type of topics and questions that are usually covered in ML specialization interviews?
Yeah that how/why distinction seems right. I’d say the regular ml design focuses more on setting it up (what are you actually solving, data, etc) and thinking through the problem to arrive at any solution that you can logically defend. The specialization is more like arriving at a good solution, and then changing it based on other stuff the interviewer adds in. Provide enough detail in both to say why your choice is better than a different choice, and if things were ideal what you’d go with. Idk your specialization, but to use a very broad ML example it would be more like why use a LSTM vs a CNN in a problem that could use either, and given what you know why did you make that choice. As far as prep, just follow the structure of that little video series the recruiter sent, and how a fb specific product would fit into that (news feed, recommendations, friends graph, sensitive content, etc)
For ML design interview, questions will be similar to practice questions you can find in grokking ml system design questions. I have no idea what to expect for ml specialization round.
thanks. appreciate your response and help.
I am preparing for MLE at meta and I al research scientist and have zero experience in system design or ML system design. How are you preparing for that ?
grokking course on both topics.
did you go through ML specialization interview? I have same on site coming
I havent yet. Do let me know if you go before me.
when is your date?
Does anyone know if you can just do two ML design interviews? Worried about passing system design since I don’t really deal with web architecture in my job :/
i dont think so. one has to be ML design, and the other system design/ML research/ML infra
Op can you share your exp
How did you do, OP?
OP please share your experience
Is this a new option? I always thought sys design was required for MLE
I had a system design, but changed to ml specialization after talking with recruiter.
Yeah I also thought sys design was mandatory for MLE. Interesting that they're doing this.