I've done about 25 ML interviews in the last 3 months. Here's my tips HM interviews are super common. KNOW YOUR WORK IN DEPTH. This is the single biggest tip I can give Be able to talk for atleast 5-10 minutes about the ML models you choose. You cant just give a high level description Understand the tradeoffs with the ML model you choose I think depth is just everything. You need to understand deeply the frameworks you use, the models you choose, the evaluation methods, deployment strategy. How it works under the hood DSA is a must. It's hard to cram DSA, but you get really really far with just a bit everyday. About 30% of interviews have a DSA round onsite. Another 30% might have DSA screening (hackerrank/codesignal). I'm not saying grind leetcode 6 hrs a day but even 1-2 problems consistently everyday adds up a lot! and you dont want to be caught unprepared if u get a DSA interview, you cant push back interviews for a month right now Takehomes/timed take homes are common. Make sure you clearly document and your code is easily reproducible and no steps are missing. e.g. if you did some data preprocessing but its not documented/shown thats bad. The conversion rate on takehomes is pretty high for me so do well on them and treat them seriously. In this interview cycle I've gotten first round interviews for all 6 takehomes i submitted Interviews are 80% luck and 20% skills in this market. Sometimes you just cant help it they might want a golang developer and even if youre a 200IQ java developer they dont care. Transferrable skills are not the same in this market. Sometimes theres just not a fit and dont beat yourself up. You can do everything right and still fail interviews Lower Pay/Lower presitge != Easier Interview. They're just different interviews looking for different things. Don't think that just because an interview is for a less prestigious company it will be easier. I've passed first rounds for roles that pay 3x more and 2 levels higher than ones ive failed. Make sure to go through the JD and note the skills theyre looking for and spend 30 mins being able to insert talking points to highlight those skills Any other tips anyone would like to add?

I'm not a machine learning engineer, so I just want to say I'm really grateful that you started this thread - Thank you! It's good to know that a lot of the interviewing fundamentals apply to MLE interviews as well (know your experience inside and out, be thorough with take-homes, be ready for DSA as it's sort of everywhere). For folks reading this who want to dive deeper into MLE, here's another great thread: How to build depth as a machine learning engineer?

[Discussion] Machine Learning Interview tips

Machine Learning Engineer at Taro Community3 months ago

Data Structures And Algorithms

Interviewing

I've done about 25 ML interviews in the last 3 months. Here's my tips

HM interviews are super common. KNOW YOUR WORK IN DEPTH. This is the single biggest tip I can give
1. Be able to talk for atleast 5-10 minutes about the ML models you choose. You cant just give a high level description
2. Understand the tradeoffs with the ML model you choose
3. I think depth is just everything. You need to understand deeply the frameworks you use, the models you choose, the evaluation methods, deployment strategy. How it works under the hood
DSA is a must. It's hard to cram DSA, but you get really really far with just a bit everyday. About 30% of interviews have a DSA round onsite. Another 30% might have DSA screening (hackerrank/codesignal). I'm not saying grind leetcode 6 hrs a day but even 1-2 problems consistently everyday adds up a lot! and you dont want to be caught unprepared if u get a DSA interview, you cant push back interviews for a month right now
Takehomes/timed take homes are common. Make sure you clearly document and your code is easily reproducible and no steps are missing. e.g. if you did some data preprocessing but its not documented/shown thats bad. The conversion rate on takehomes is pretty high for me so do well on them and treat them seriously. In this interview cycle I've gotten first round interviews for all 6 takehomes i submitted
Interviews are 80% luck and 20% skills in this market. Sometimes you just cant help it they might want a golang developer and even if youre a 200IQ java developer they dont care. Transferrable skills are not the same in this market. Sometimes theres just not a fit and dont beat yourself up. You can do everything right and still fail interviews
Lower Pay/Lower presitge != Easier Interview. They're just different interviews looking for different things. Don't think that just because an interview is for a less prestigious company it will be easier. I've passed first rounds for roles that pay 3x more and 2 levels higher than ones ive failed.
Make sure to go through the JD and note the skills theyre looking for and spend 30 mins being able to insert talking points to highlight those skills

Any other tips anyone would like to add?

341341 Views

33 Comments

Discussion

(3 comments)

4
Alex Chiou
•Tech Lead @ Robinhood, Meta, Course Hero
3 months ago
I'm not a machine learning engineer, so I just want to say I'm really grateful that you started this thread - Thank you! It's good to know that a lot of the interviewing fundamentals apply to MLE interviews as well (know your experience inside and out, be thorough with take-homes, be ready for DSA as it's sort of everywhere).

For folks reading this who want to dive deeper into MLE, here's another great thread: "How to build depth as a machine learning engineer?"
4
Rahul Pandey
•Tech Lead/Manager at Meta, Pinterest, Kosei
3 months ago
Such valuable insights here! One thing I want to comment on:

Takehomes/timed take homes are common

Do you think this is more common for ML interviews compared to normal SWE roles?

I don't like doing take-home tests, so my general advice is that you can simply ask the recruiter to have a technical phone screen instead. You'll be surprised how often you get what you want simply by asking.
- 1
  Machine Learning Engineer [OP]
  •Taro Community
  3 months ago
  tbh im not sure as I havent done a ton of SWE interviews. I will say that the key difference is that ML interviews have multiple rounds but each interview is a different round. Can be any combination of -- SQL, System Design, ML Theory, Pair programming, DSA, case study, behavioral
  
  To be honest I'm not really good at leetcode and I was able to get by without leetcode so far as a junior so I would likely have failed if it was a live DSA round. Both the on-sites were directly after completing the OA/take home so they did treat it similar to a first round
  
  The take home tests were generally similar to what you'll be doing on the job so you dont need to prepare for it at all. I passed my take homes without prep
  
  But the take home took me 2-4 hours per company and I preferred the take home but its definitely good to know that you can ask for a technical round instead!!