Original link: https://muyun.work/2022work.html
1. Determination of position, track, enterprise and work place
1.1 Determination of job-seeking positions
The primary goal of the job search preparation period is to determine the position !
Generally speaking, it is divided into technical posts and non-technical posts, and there are various branches under it
- Technical posts include: research and development, algorithms, and testing. R&D includes front-end, back-end, client, database, operating system, cloud computing, games, etc.; algorithm positions include vision, voice, text, search, promotion, planning, etc.; I don’t know much about the subdivision of testing.
- Non-technical posts: product manager / public examination / clerical staff, etc.
Please respect everyone’s choice. Students majoring in computer science can also show their own advantages in non-technical positions. The determination of the position here is not to rule out other possibilities, but to increase the priority of a certain item. And do the elimination method in this process, and after trial and error, you can recognize yourself more clearly and narrow down your options.
1.2 Investigation and exclusion to determine the remaining dimensions
Before or after determining the position, it is necessary to read a lot of JD (Job Description, job description) to eliminate information barriers. The specific method is: collect the JDs of various companies (listed companies, state-owned enterprise research institutes, unicorns, etc.), and find out their public requirements and special requirements. For example, R&D will require programming languages: Golang, C/C++, Java, etc., and there may be others, such as programming experience under Linux, network programming, etc. After sorting out a table and sorting out all the job descriptions, you can roughly understand the tendency of the current employment environment, and you can also calculate the greatest intersection with your own skill tree, check for omissions and fill in vacancies to find a job that suits you. In the field of investment, hundreds of industry terms are enough to eliminate the information barriers of the industry, so after reading a lot of JD, you can also understand the industry jargon during job hunting, and understand the requirements of the skill tree better, and if it is common Content, the terms with the highest frequency are the most important.
On this basis, you can further do the elimination method to make your job search portrait clearer: the track, the company and the place of work. What track and industry do you work in? Aerospace, autonomous driving, medical AI, AIGC, etc. Each track and industry has its own characteristics, and you need to do a good job of research. In terms of enterprises, ask yourself your own inclination, whether it is a state-owned enterprise research institute or the Internet, or a unicorn or a start-up company. This involves work-life balance, salary treatment, personal growth and other perspectives. Know yourself to make the right decision. Finally, there is the work location. Some students will determine the location quickly, but some students will delay the decision until the last moment. Personally, I have been vacillating between the two options of going back to Sichuan for development and going outside for a few years, and finally decided to temporarily Leaving the comfort zone and looking for opportunities in the wider outside world, after all, there is no such thing as Sichuan.
Everyone has their own reasons to determine these four dimensions. These questions may change in various ways before applying for a job, but I think it is better to do the elimination method. First draw a circle and keep narrowing the circle. You can also gradually realize what you want.
1.3 Personal job search ideas
My personal main ideas for job hunting are as follows: position: computer vision algorithm engineer; enterprise: escape from the Internet and embrace unicorns; track: automatic driving; location: Chengdu/Yangtze River Delta. My reason for fleeing from the Internet is: I have seen the algorithm of Meituan to squeeze delivery people, the recommendation algorithm of Douyin, the algorithm of Weibo and WeChat Ant’s risk control and content review. So the technology of Internet companies may be top-notch, but I think if I make a product that I don’t like/values I don’t agree with, I don’t think I will be happy either. The following also includes my personal thoughts on geographical selection:
Thoughts on Individual’s Regional Selection
It is difficult to take the first step of job hunting, because it takes a lot of effort to determine these dimensions. I hope that you can recognize yourself more during your studies, talk to yourself more, and find out what you really want in your heart. Once determined, do it without hesitation
2. Resume and interview preparation
First of all, we will apply for a job in two processes: submit your resume -> interview (subsequent offer selection, salary negotiation and other operations are not covered in this article)
Then our goal is actually divided into two parts: screening through resumes and screening through interviews
My personal situation is China Southern Airlines undergraduate and master, and I only did job hunting in the direction of algorithm post, so my experience may also be more applicable to students who are in the master’s stage and have similar job hunting directions. Next, I will focus on target disassembly and optimization ideas from the perspective of resume and interview.
2.1 The resume of the algorithm post has three axes
In my opinion, the algorithm job resume has three axes: scientific research papers, project experience, and internship experience
2.1.1 Research Papers
Master’s students should more or less produce scientific research papers, and it should be said that this is the main achievement during the master’s period. In recent years, the increase in the number of manuscripts accepted by top conferences in the field of AI algorithms has also created serious inflation of papers. There are rumors on Zhihu that there is no algorithm for top conferences, and even one top conference cannot pass the resume screening and the like. argument. But we must keep in mind: correlation does not equal causation. It is not because there is only one top-level paper that cannot pass the resume. A large part of the reason for the failure of the resume is that the resume does not match the position. No reason for screening. In terms of scientific research papers, there is no way to speed it up, only rely on usual hard work and diligence.
2.1.2 Project experience
Note: The projects mentioned in this article include school-enterprise cooperation projects, algorithm competitions, and open source projects.
School-enterprise cooperation project : This is highly related to the research group. It depends on whether the tutor is connected to the horizontal and whether the horizontal is assigned to you. It can be a pain to do, but the pain you will enjoy while writing your resume is worth it. If you don’t have much confidence in your resume, doing some school-enterprise cooperation projects will also lay a solid foundation for yourself, and may produce ideas for thesis, which can serve multiple purposes.
Algorithm competition : As more and more companies integrate algorithms into their products, more and more companies continue to launch algorithm competitions (here specifically refers to computer vision and natural language processing, excluding ACM) and set up bonuses to set up everyone’s algorithm solutions. The cycle of the algorithm competition is usually three months to half a year, which is suitable for students who like to practice and get positive feedback on the competition rankings.
Open source projects : The trend in recent years is that interviewers have seen too many projects described on their resumes. Most of the projects have no gold content, and the details cannot withstand scrutiny. And if you are a Contributor of an open source project, the interviewer will have a good impression of you most of the time, because you have some valuable PRs that have been incorporated, which can reflect at least a few things: you are familiar with the use of Git; Understand mid-to-high-level and complex projects and find bugs and fixes; the code style is acceptable; the self-driving force to write code is good, etc. Open source projects are actually not long, and there is no clear node. And you can find a lot of clear problems in the issue, and the maintainers will also throw out some problems to be fixed, eager for community students to contribute, without having to work hard to think of ideas like scientific research. Many advantages make me also highly recommend that everyone contribute to open source projects to enhance their competitiveness. For further reading, you can read Teacher Liu Weipeng’s article: How to spend two years interviewing a person
How to participate in open source projects, here are some open source project platforms:
- Google’s GSoC: https://summerofcode.withgoogle.com/
- Alibaba’s Alibaba Summer of Code: https://asoc2022.opensource.alibaba.com/
- CCF’s GitLink Programming Summer Camp: https://www.gitlink.org.cn/glcc
- The Open Source Summer of the Institute of Software, Chinese Academy of Sciences: https://summer-ospp.ac.cn/2022
- OpenMMLab’s Super Video Camp: https://openmmlab.com/activity/codecamp
- PaddlePaddle’s Flying Paddle Hackathon: https://aistudio.baidu.com/aistudio/competition/detail/776/0/introduction
1.1.3 Internship experience
When submitting the internship, most of the students had no internship experience, including me. Therefore, you can leave it empty when submitting the internship, and in the autumn recruitment, if you can have a high-quality internship experience and be good at summarizing, it will greatly increase the success rate of the interview.
1.1.4 Resume Preparation Advice
You should prepare these three tricks according to the actual situation of your own research group and individual. Judging from the market situation in the past few years, students who are particularly good in one aspect can be very popular in the talent market, such as competition masters/Kaggle Masters, academic masters who hold multiple top conferences, and multiple high-quality internships. Big brother; but more students can’t be so focused, so those who can do well in two aspects can also have a good place. And if it is a triangle fighter, you must highlight your relative strengths, and focus on two of them.
So how should you prepare? For the algorithm post, the most competitive is top-level papers. If the group can produce top-level papers, it can devote more time to scientific research. But this kind of group is a minority after all. If the situation of the papers in the group is worrying, then the first thing to do is to adjust the focus to practice and projects. The precondition for practice is that the supervisor will let people go out for practice. It is recommended to communicate with the supervisor in advance. What are the conditions to go out for an internship (our group basically completes two works, and the internship can be done without delaying the graduation thesis), so as to plan your own time node.
In terms of resume writing, be sure to use the STAR rule to write a thesis. My personal favorite sentences are:
- This project/paper/competition aims to solve the XX problem and we propose the XX algorithm framework. The content of personal responsibility is: XXX. In the end we achieved XX results (reaching SOTA performance/improving algorithm efficiency/obtaining competition rankings)
2.2 Three tricks for the interview of the algorithm post
In my opinion, the three tricks of the algorithm job interview are: communication skills, basic skills, insight skills;
1.2.1 Communication skills
First and foremost is the ability to communicate, including self-introduction and a brief introduction to one’s past experience, as well as answers to questions about experience details.
1.2.2 Basic Capabilities
The basic ability includes the inspection of code ability and domain knowledge.
The ability to code means that the interviewer will ask some questions that need to be implemented by programming, some of which are algorithm questions, similar to Leetcode, which provide input and output samples, let you write code, and have time complexity and space complexity requirements. Another category is post-related code implementation. Taking the AI algorithm post as an example, SenseTime likes to ask you to write a BN, cross-entropy function, residual module, evaluation index, etc. by hand. I myself have also encountered the producer-consumer model (the knowledge category of the operating system) that made me write it by hand.
The investigation of domain knowledge generally includes general knowledge and job-related knowledge, which we commonly call stereotyped essays, because most of them can be memorized by reciting and understanding. Taking the algorithm post as an example, I will ask basic knowledge such as the principle of BN, convolution parameter calculation, loss function, etc., and I will also ask some post-related questions in combination with the post, such as the development process of the target detection algorithm, anchor-free and anchor-base, transformer , self-monitoring, and more. Sometimes the interviewer will also check your familiarity with your research field based on the content of your resume.
1.2.3 Insights
The section on insight ability is relatively broad and includes a lot of content. Here are a few examples
Example 1: Give a solution to a problem in a certain scene. When I was on the side of the byte, I gave such a question: How to identify the corresponding product from the live broadcast room when Douyin live broadcasts bring goods.
Example 2: Ask your opinion on a certain field, open-ended question and answer. Talk about your understanding of Transformer; why do you think generative self-supervised algorithms can work.
Example 3: Inquiry about programming habits or research habits. The director of Huawei Noah was asked this question: who are the senior scholars and groups in your research field; what do you think is the most important thing to do research
The first example is a very common situation. The interviewer may ask you about what he is doing. It can be understood as a free prostitution scheme, but it can indeed screen out high-quality candidates.
1.2.4 Interview preparation advice
- Self-introduction is the same as writing a resume, you must follow the STAR rule to tell your past experience. You must be familiar enough with the details of the experience to avoid vague descriptions that will cause the interviewer’s distrust. In addition, it is also important to guide the interviewer in a familiar direction. For example, when there is a deviation in your understanding of your experience, you can answer him first and then explain, “Actually, my research work during school focused more on xxx, and I think xxx”. In addition, communication skills are also reflected in some details. For example, when he gives you an algorithm question, if you come up with an idea, you need to communicate with the interviewer about the correctness of the idea first. The interviewer who is more nice will give you some hints; If you have no ideas, you can also talk about your general ideas to avoid embarrassing situations. There are two better ways to cultivate communication skills: content creation and output on public platforms; and mock interviews (Mock Interviews) with small partners in similar directions to check for gaps and fill in gaps, and gradually get used to the state of interviews.
- In terms of coding ability, I personally don’t have so many questions, about 150 or so. The process of brushing the questions is to follow the code caprice to finish most of the questions. After mastering the general ideas and code templates, I start to brush codetop, so the platform I use for brushing the questions is only code caprice and codetop. Facts have proved that as long as you master the high-frequency questions of codetop, you will most likely encounter the original questions or modified questions. The most important point in the process of brushing the questions is to remember to review. I will open a GitHub warehouse to record my own process of brushing the questions. Every time I make a mistake, I will write it in the form of a note ( link to the personal warehouse ). I will mainly review the old topics that I don’t know, and the new topics are relatively limited. In the end, some topics will even be written seven or eight times. This is my stupid method, but I personally feel that it conforms to the Ebbinghaus forgetting curve and has strong applicability. As for programming specifications, you can refer to the code style guidelines of major manufacturers, such as “Google Open Source Project Style Guide”; you can also do personal open source projects and contribute code to existing open source projects.
- As for stereotyped essays, open a document by yourself and keep summarizing Niu Ke’s face-to-face experience (you can add it together with your job-seekers), and continue to expand the boundary of a knowledge in the form of questions until you fully understand it. By point to point, you can actually predict what questions the interviewer will ask you. I will read Teacher Jiang Dabai’s “Artificial Intelligence Algorithm Gang Jianghu Wulin Cheats”, which contains many questions that can be used to check for omissions and fill in vacancies.
- As for the insight ability, it is unlikely to be fully prepared through advance preparation, because it is impossible to predict what scene questions the interviewer will ask. But some more common scenarios can accumulate solutions by themselves, and then you can lean on them when analyzing problems. For example, the optimization of small target detection and segmentation, how to detect noise in labels, etc. In addition, you can pay more attention to the technical public accounts of some companies, and some high-level technical solutions will be given, but the technical details will be relatively few, and you need to collect further by yourself. For example, Meituan technology, Dewu technology, iQiyi technology, Tencent technology, Xiaohongshu technology, Kuaishou audio and video technology, etc.
Brush algorithm questions and organize eight-part essays
2.3 Preparation tips
In the job hunting preparation stage, we must get rid of the inertia of thinking. Recording knowledge in the process of constantly understanding terminology is just input; stringing knowledge together to output opinions is knowledge internalization. My personal habit is to use the card note-taking method to record knowledge, gather fragmented knowledge and output in one place, and review and summarize from time to time. In addition, we must actively obtain evaluation and feedback : no matter whether the instructor allows the internship or not, we must bravely submit resumes and interviews, and grow in the process of being continuously evaluated. The following is my information channel sharing and information collation practice, for reference
- “Meditations on Products”: Measure yourself rationally, immersedly, and objectively, and draw nourishment from those who calm down
- RSS + RSSHub: Let the information stream be actively input, https://www.bmpi.dev/self/use-rss-email-read/
- China Southern Airlines Blog Collection Plan: https://blogroll.a2os.club/
Focused information channels 1
Information Channels of Concern 2
Text output and summary
Thoughts during the autumn recruiting period
3. Autumn Recruitment Actual Combat
After the first two stages, after keeping thinking and successfully practicing, I have a deeper understanding of myself and my personal judgment will be more accurate. You need to define what Offer is satisfactory. What we need to do at this stage is to adjust our mentality: know yourself and adjust your expectations
- Look at more similar cases and listen to your inner voice
- Feel more about the current job market situation, even if you adjust your expectations (the job market in 22 years is deserted, everyone must stick to it)
- When making a choice, seek advice from others: family, friends and seniors
Finally, talk about salary and selection. I don’t have much experience in this part. I even forgot the A-point option because I waited too long and was too excited during the lottery. So I would like to give you two suggestions:
- Have the courage to negotiate salary and wait for your favorite company
- If you still do not have a satisfactory offer, write a paper and actively prepare for spring recruitment. Don’t overestimate your tolerance for less-than-ideal jobs
Finally, a little tip:
- You can pay more attention to the official employment accounts of other schools. For example, if you want to go back to work in Sichuan, then follow the employment official account of Sichuan University Chengdian. If you want to go to the Yangtze River Delta, then follow the employment official account of Nanjing University, Southeast University, Zhejiang University, Shanghai Jiaotong University, Fudan University, Tongji University, etc., and you can even join them job fair. When applying for a job, recklessness is an advantage and a manifestation of bravery.
4. Message
Postgraduate life is not limited to the resume, I did not write a lot of content on the resume. During my master’s degree, I participated in the school research meeting and met some good friends. I joined the school’s choir and performed at the Shanghai Poly Grand Theater. These constitute the bright spots in my postgraduate life, and the communication with my classmates has further clarified my job search direction. There’s no denying that we all have a preference for “grand narratives.” But in fact, the success of a grand narrative often comes from the most important one or two points, and for these two points, someone must pay attention to the details from beginning to end, otherwise it will be a bunch of leaks everywhere. A castle in the air.
During the job hunting season during this period, most of the students around me were not as happy as they wanted. This anxiety also spread to the next class of students. They were like the next classmate who was about to be injected, unable to laugh. This year’s autumn recruiting time is very long. From summer to winter, the cold air is very real to us. Everyone must persevere. I had a nosebleed in the sharing of the preparation department seminar at noon, but fortunately, the sharing in the afternoon was more successful. On the way home from leave, I bought fruit and found a rainbow in the sky. The interviewer who interviewed at night was very nice, and the next morning, I received an email of being approved as a full-time employee. good news keeps coming
It is going uphill now, and it will be very hard to walk, but in fact, it will be very, very fast during this period of time. I am rather stupid. In the past days, I actually did not walk very fast. I am very pleased: I am walking slowly, trying different directions and outputting words, recording what I think,
I gradually formed a stable thinking paradigm, and these countless past experiences gave me confidence: problems can always be solved, as long as I keep going. It also confirms the view that slow is fast. The hymn of human beings is the hymn of courage. I have been mentioning that we must have courage, step out of the comfort zone of thinking, submit resumes and interviews, wait for the company we like, and so on.
“Life was too much, and it was difficult in every way at the time. Looking back one day, it turns out that I have flown across thousands of mountains.” This is what my freshman counselor will say to us when I meet. I always feel very powerful and share it with you. I hope that everyone’s efforts will be rewarded, the postgraduate life will be happy, and the autumn recruitment will go smoothly. Thank you!
5. Reference materials (recommended to read)
- Recommended Algorithm Going Ashore—Preparing for Internship Autumn Recruitment
- NUAA-INTERVIEW: https://github.com/NUAA-Open-Source/A2OS-INTERVIEW
- Use stochastic gradient descent to optimize life (frequently read and often new): https://zhuanlan.zhihu.com/p/414009313
- The painstakingly organized guide after two school recruitments (there are many introductions of job hunting slang): https://sspai.com/post/64458
- How to spend two years interviewing a person (explains the importance of GitHub open source projects from the perspective of the interviewer): http://mindhacks.cn/2011/11/04/how-to-interview-a-person-for- two-years/
- Thinking of internship and campus recruitment (understanding autumn recruitment from the perspective of strength and potential): https://zgq.ink/posts/internship-and-campus-recruitment
- School recruitment interview experience written in early 20s, self-study CS experience and job sharing (interview skills and preparation process, very detailed): https://github.com/conanhujinming/tips_for_interview/blob/master/README-zh_CN.md
- 【Methodology】School Recruitment Sui Sui Nian (Introduction to Autumn Recruitment Methodology): https://imageslr.com/2021/autumn-recruit.html
- An Incomplete Guide to Campus Recruitment Interviews (the author’s other blogs are also very interesting): https://www.zackwu.com/posts/2020-10-05-an-incomplete-guide-to-campus-recruitment-interviews/
- Do you have any questions about internships? : https://0xffff.one/d/940-wen-yi-xia-ji-ge-guan-yu-shi-xi-de
- In 202x, where will the school recruit algorithm jobs go: https://mp.weixin.qq.com/s/kt3TBoQiL0SH_6XjjCoAmw
- Recommended Algorithm Landing Process – Preparing for Internship Autumn Recruitment: Recommended Algorithm Landing Process – Preparing for Internship Autumn Recruitment
- Some suggestions for recruiting students in algorithm schools: Viewpoints: https://zhuanlan.zhihu.com/p/567878099
- Algorithm Engineer’s “Between Heaven and Earth”: https://mp.weixin.qq.com/s/XV76d-c9LRA_VCOSO1Bx0g
- Genius list: https://zhuanlan.zhihu.com/p/475366689
This article is transferred from: https://muyun.work/2022work.html
This site is only for collection, and the copyright belongs to the original author.