As an AI Researcher in Liulishuo

by Kaili Zhu

This post is about casual internship life, not professional skills.

After my second time finishing the great course Stanford CS231n, I was curious and anxious to know what I could do about Deep Learning, since I did not have someone to study with, or some specific project to work with. Then, I decided to get an internship about Deep Learning somewhere and searched for such opportunities. Luckily, I saw the recruitment information of a good company that I heard before, yes, it was Liulishuo (official name: Yuguan Information Technology (Shanghai) LLC), so I plucked up my courage to submit a resume to it, and more luckily, the HR transfered my resume to a colleague (my future supervisor) even though she checked the resume and noticed that I did not have any practical experience in Deep Learning before. Then I passed the interview and became a member of the algorithm department of Liulishuo.

In Liulishuo, I really learn a lot! I started to know what Seq2Seq Model is, what kind of Attention can be used, how we can apply neural networks to solve Grammatical Error Correction problems and so on. My main job is to optimize our model, so I have the opportunity to search for papers, read and select them, implement some key technology into our model using TensorFlow, run my code with GPU and wait for the results, analyse the results and make reports. In this process, I can always update my knowledge base and use the advanced technology to improve my work, so I benefit a lot.

What I do in Liulishuo is interesting and impressive, the colleagues I work with there are also excellent.

My supervisor Ruobing Li, is an expert on NLP. He can tell what the reason is when we propose a good model or a bad one. Much of my research and work is under his guidance, and I am very grateful.

Another colleague I mainly work with is Chuan Wang. He is a creative person and can come up with good ideas to improve our model. My initial work is based on his latest paper, and we made a paper together this early year.

The Chef Scientist of Liulishuo, Hui Lin, is an excellent researcher in machine learning, and you can find this in his Google Scholar page. I have met and had meetings with him several times, and it is always helpful to talk to such a person who know a lot deeper and broader than me.

Liulishuo has an AI lab in Silicon Valley, and there are some outstanding staffs who can post papers to the top-level conferences. I have contacted with the leader of that AI lab, Prof. Yang Liu, when I was writing my paper, and she helped me a lot. One of the most impressive things is, she can modify a whole LaTeX without running it, and the final version is nearly without any mistakes or errors. For me, it sounds like writing code in a document whiout IDE to check the syntax errors, and we can run the final code perfectly.

Last but not least, there are other points that attract me to stay in Liulishuo:

  • a competitive internship salary
  • a Macbook Pro with top configuration to work with
  • free lunch and dinner (only in Shanghai office, not in Hangzhou office)
  • free snacks (also only in Shanghai office)
  • foods on TGIF (Thank God It's Friday, Shanghai only, sigh)
  • ...

Another important point I would like to stress on is the flexible working place. I can work from home if I am inconvenient someday, and it is hommization.

As a fast-growing startup, Liulishuo has some inevitable shortcomings. For me, the biggest problem is the poor VPN to connect my GPU machine, because the tunnel is always crowded and I can not operate on my server well. So, if you are a good network administrator, welcome to Liulishuo and help us to solve it :).

I have been working in a stable big company as NetEase, and in a fast-growing startup as Liulishuo. For me, to choose what kind of company to work in is secondary, what matters is, what kind of job I can do or I am expected to do. I enjoy the process to learn, to do research, and to pursue the state-of-the-art. So a company that aims to be the best one in a promising field, encourages its employees to keep going and achieve the state-of-the-art, also with flexible working time and place, like Liulishuo, will fit me perfectly.