My deep learning job interview experience sharing
Recently, I’m seeking a job about machine learning or deep learning, and until today, I’ve interviewed 6 companies, maybe there are more (hopely need no more). Maybe my experience can help people like me, so I decide to write it down and share with you.
Let me introduce myself briefly, my name is Justin, I’m from guangzhou,China, and interviewing some Chinese companies. I’m not a CS student, and graduated in 2015, but I started learning machine learning and deep learning myself until today . I think I’m ready to enter the AI industry, so I quitted my previous job in February,2017. Ok, let’s check out my experience.
1st company: aim at market research and information processing
This company don’t have deep learing team yet, they want to build a team to do some information processing, but they don’t own these data, they just buy the data or download from the internet. After access these data, use machine learning to process it and sell the result to other companies.
I don’t think they will do a great job, since MLaaS have no future, and what’s worse they don’t create their own data, so I’m not interseted in this company.
The interviewer is CTO, he asked something about my previous job at first, and then asked if I can use tensorflow. He didn’t test my capabilities in machine learning or deep learning, just gave a picture about what’s they gonna do with deep learning. I think the CTO didn’t know much about ML or DL, just want to build a team. But through the conversation, he gave me a sign that the conpany can’t pay to much to hire me, and what’s more, after 2 days later, he called me and said: oh, your capabilities doesn’t match our job very well, just match part of it, let’s talk face to face to discuss your job and salary, if you are busy these days, just take your time, we are not hurry to hire you！
To be honest, I’m not so comfortable after hear that.
2nd company: aim at crops disease detection using computer vision
This company has a big paddy, their staffs check the crop every day to see if the crops is suffering disease or insect pest. I’m very intersted in what they about to do, since there’s few AI application in agriculture. Besides, they have their own data (paddy). What they want to do is use camera to take photos for crops, and use CV to detect if there’s pest on it or if they’re turing yellow.
But the big question is, they don’t have any tech teams! Even they have no website engineer! Therefore the interviewer can’t ask me any technical question, she just seeking my opinion about how to achieve their goals and really appreciate what I can do.
But she didn’t call back again.
3rd company: Automated creativity
The company is a SaaS company, one of their service is providing a lots of picture elements for user to create their own new poster or website banner. The most awsome fuction is, the system can generate a new picture according to a single discription sentense provided by user! That’s the oppisite side of image captioning, the interviewer said they have already created a demo verison, but somehow I don’t think so, according to our whole conversation.
The interviewer is a girl, her age is similar to me, I can tell she’s major in CS, that’s because she always test my knowledge about data structures and algorithms! And didn’t ask any question about CNN or computer vision! The only one question about machine learning is the algorithm principle about k-means (thank god she finally tests my ML knowledge), I can’t answer her data structures and algorithms questions very well, since I’m not a CS student, but maybe I’ll learn that later.
Finally, she said they want an expert in CV deep learning, 5 years working experience and PhD better. Perhaps they are looking for Google, Facebook engineer?
4th company : aim at internet and big data safety
This company I didn’t know much, but I think they provide internet safety service to goverment or other company. The manager said there are about 6T data go though their equiments per second! With such big amount of data, they want to dig something useful form it, so they need the machine learning team to do it.
There were 3 interviewers, one is the manager, he was just sitting there and hearing our talk. The other 2 interviewers are the ML team members, basically I talk with them. The girl who talk to me most is a master in CS, she asked me about some machine learning algorithm, such as k-means, random-forest, decision-tree, but seldom asked me about deep learning. She said that the team don’t use deep learning for now, but they will consider apply deep larning later.
Actually we have a great talk at that time, I showed my knowledge in deep learning, and we exchange some views about the what they do. It’s a pity that I didn’t recieve their call later, maybe they need tradictional ML engineer, or maybe I didn’t remember some ML algorithm details and make them think I’m not good at that. But it reminds me to review those ML algorithms.
5th company : financing company
This financing company want to recruit NLP engineer, although I’m major in CV, they still invited me to have a talk. The interviewer is professional, probably a CS master, he test me a lot in math, let me do some formula derivation, for example, give me a simple linear model, let me write down the explicit solution of W and b. Forgive me I don’t know how to do that for now, but I’m still learning the formula derivation.
And then, he asked some questions, some of them are interesting:
- We all know that we shouldn’t set the initial weights to zeros, but under what circumstances, we can set them to zeros?
- What’s the mathematical priciple of momentum, RMSprop, and Adam?
- What’s the difference between those imagenet winner models?
- Would you eager to implement the technique from papers? And if you do, do you think your math level can handle that?
- What’s the usage of embedding layer? And why we need this layer?
In a word, the interviewer would like to test my math knowledge, I admit that I’m not doing quite well for now, but I’ll keep learning.
6th company : electronic commerce company like amazon
This company has large amount of data, such as emails, pictures that seller upload to their server everyday. Their AI department want to detect the pictures and do some picture processing automatically, so that the company can save much human cost to do the picture porcessing.
The interviewer is pragmatist, he let me introduce some algorithm detail that I most fimilar with, that’s no problem. And then, he asked me the intuition about CNN, backprop, SGD, and what’s the difference between keras and theano , I can handle that.
One thing he impressed me is that he won’t ask me to use only one or two tools to build the model, as long as I can solve the problem, no matter which framework you use or whatever coding language you use, that’s all ok!
In a word, I think that’s the best interview I’ve ever had, seems the interviewer is satisfied by my behaviour, maybe I can get a luck that day!
That’s all the interview experience for now, if I have following interview, I will share with you guys, and if you have something want to talk about your experience, we’d like to hear, hope our experience can help each other! That’s my first blog in medium, I’ll write more in the following days, please give me a like or follow!
