March 13, 2026 • 12038 words

Applying for Grad School in the US

I am thinking about applying for masters in CS in the US. It's already February. I thought about applying before but my mind was dominated by stress and the anxiety of failing classes. I applied many times before, but this is the first time that I am in a such a position. Ok, this is the first time perhaps I am aiming for a non-top-tier school in my life. I am not applying for the top 10 or 20. And any excuses are just futile. Yeah, it's February 10th now. I am late. I am going to try to apply for fall, if it doesn't work, I am going to try to apply for spring. I don't want to wait for too long.

School Tiers

I am applying for research masters in ML/Data Science now and eventually, I want to get a PhD. First I need to know the general tiers of universities in the US. I thought there were only 400 universities in the US when I was looking at USNews. But it turned out like there are thousands of institutions in the US offering a higher education.

CSRankings

If you go on CSRankings and filter for AI in the last 3 years, you see China taking the top 6: Tsinghua, Zhejiang, Peking, SJTU, Nanjing, and USTC, followed by 2 Singaporen universities and a Korean university, while the US is out of the top 10. Change the filter to the last year, and China takes the top 8. Chna takes more than half in the top 20 in "AI" in the recent years on CSRankings. This is so surprising that my university is literally #5 or #6 on that list not only in China, but in the world. Yet on the ground reality is that the top students from Tsinghua, let alone USTC or Jiaotong University would fight for a PhD in the "Big 4", and most would glady take an offer from schools such as UCSD, UCLA. People vote with their feet to go to the US. So basically CSRankings is highly unreliable for the absolute rankings, you can use it to search for professors but you shouldn't use it to rank US schools as well.

On the other hand, if the school appears on the list in CSRankings, it usually means there is a functional PhD program, vice versa. I went on CSRankings and searched for the last 3 years and to my surprise, only 150-200 schools in the US appeared on the CSRankings at all. Most of them are R1, around 10-20 are R2 (some with few faculties on CSRankings).

Basic Picture

Anyway, here are the basic rankings. It all depends on the specific lab and the specific professor, and everything is like "roughly" or treated as a guideline. We clearly can't just copy CSRankings.

  • Top (big 4): Stanford, Berkeley, MIT, CMU
  • Tier 1.5: UCSD, UCLA, UW, UIUC, UMass Amherst, UMich, Maryland, etc
  • Tier 2: UCI, UCSB, USC, Virginia, Chicago, Rutgers, Ohio State, etc
  • Tier 3, solid R1 (50-100): UCR, UCSC, Oregon State, George Mason, Colorado Boulder, WashU, Tufts, etc. They do serious research but aren't considered "prestigious", though it all depends on the specific lab and professor.
  • Low R1/High R2 (100-150): DePaul, Portland State, Ole Miss, Idaho, Boise, Mizzou, Kansas, West Virignia, Iowa, Hawaii at Manoa, etc. Appears on CSRankings, has a basically functional CS PhD but few professors, maybe 1-3 per field, or they have a very specialized field.
  • Unranked/Teaching Universities/CCs: San Jose State, Cal State, Wake Forest, James Madison, Pace, Miami of Ohio, Alaska Fairbanks, etc. No CS research, basically no CS PhDs. You either prepare to get feed directly into the industry, or simply pay to enjoy your life in a school with a funny name. I don't want to be here.

Geography

I've been to US 4 times now

  • California and New York (1st grade winter)
  • Great lake area and Boston (6th grade summer)
  • So-cal and Irvine (7th grade summer)
  • Berkeley and Alaska (sophomore exchange year)

So in my mind US has 3 main parts:

  • New York and New England ("BSC" in Connecticut, "In the Unlikely Event" in New Jersey, "Wonder" in New York, "Magic Tree House" in Pennsylvania)
  • Silicon Valley and California ("La La Land", "The Graduate", "Easy A", "The Joy Luck Club")
  • Seattle and Portland ("Twilight" in Forks)

Then there are 3 smaller but still significant areas:

  • Florida (hot)
  • Texas (hot)
  • Chicago and Michigan area

Then there are the "middle of nowhere" states:

  • Kansas (Wizard of Oz)
  • Missouri (Tom Sawyer)
  • Utah (Mormons)
  • North Carolina ("Along for the Ride")
  • West Virginia (Mountain Mama)
  • Alabama (Atticus Finch)
  • Mississippi (Roll of Thunder)
  • Louisiana (Jambalaya)
  • Alaska (Call of the Wild)
  • Hawaii (Aloha)
  • Puerto Rico (Despacito)

I don't know much about the rest.

My Background

  • TOEFL: 109(4 years ago, 16)/105(2 years ago, 18)/104(half a year ago, 20)
  • GRE: 155(Verbal)/166(Quant) (4 years ago)

So I am in USTC, studying CS. My overall GPA is 2.7/4.3.

Are you happy? You can laugh at me, it's funny, even if I am crying or doubting my sanity. I don't know if I should blame it on myself or the Chinese CS education. Because here are the courses I took at Cal with literally zero CS foundation in one year I took.

  • CS161 Security (A)
  • CS170 Algorithms (A)
  • EECS149 Embedded (B)
  • CS188 AI (B+)
  • CS182 DL (B+)
  • CS285 RL (C+)
  • CS61C Architecture (C+)

2 math courses

  • Probability (P, I slipped in the midterm)
  • Abstract Algebra (A)

And I came back to USTC, and I still needed to complete the standard CS curriculum in China, and I got slaughtered. And I am finishing my undergrad this year!

I had no research so far. I was too busy dealing with depression and my mental health after the sophomore year. Good news is that I do not have suicidal thoughts right now. My weight went up in a year from 72 kg to 86 kg. Then I went out walking 10-20 km on desolate streets from midnight to 5 am. My parents were first "very disappointed with me" and completely shocked how "I was fine in Berkeley and we thought everything ahead was cleared". Then last year in May there were "we are graduating" signs in the school and I felt my heart wrench every time I looked at them. I learned Russian largely entirely because of the depression when I would watch YouTube all day as a form of escape. Looking in retrospect, it's divided into 2 phases, the first 2 years I almost didn't know anything about CS, then I was struggling/depressed. I don't blame myself anymore, but once I get my feet inside the master's program, I need to try to write and publish papers. I am very lucky (I am going to graduate from USTC soon after passing the failed courses), optimistic (things would be better almost certainly) and resilient of failure!

I am not dying for prestige anymore. In short, I am likely paying to get into a tier-3 program. I do like math and CS, but I hate the utilitarian way how most students treat them. I have a romantic, poetic soul. Art, writing, and reading literature are not negligible in my life. This is my valid entertainment and relaxation as I do not game or drink, and dropping it for productivity only results in burnout or depression. I already spent so many years in a sterile environment and the CS/DS department in the US is also going to be quite utilitarian, filled with Asians. I must not spend my Masters and PhD in a sterile university.

Fatalism

掩重门、浅醉闲眠。莫开帘，怕见飞花，怕听啼鹃。

(高阳台·西湖春感, 张炎)

The annoying thing is that you think you did nothing wrong, you followed your logic, you worked normally, you tried, and you still failed spectacularly. In those scenarios you began to doubt your sanity, nihilisim kicks in, and you get fatalistic, and you are unsure about what to do in the future either. Prestige and grades matter less because there's no control over them, while the process of learning and the quality of life begin to matter far more. I mean, this is not the first time that I feel fatalistic. Back in my high school, one of the top in Shanghai but a "Gaokao machine", I was getting extremely poor grades on every subject except math and physics that I was almost doomed to fail in Gaokao and not get into a "985". Yeah, even English, when I was reading full-length English novels and watching YouTube without problems, getting a good score in TOEFL. I was always one of the worst in exams. People didn't believe it because I went to USTC at 16, but I went in through the special program where I only needed to take math and physics, and meet a low baseline in Gaokao, which I just managed to reach the exact cutoff line. One point less in Gaokao, or one point less in the specialized exam, I would be dead. But then I failed so hard in USTC ironically. Although for the exchange year in Cal, that was less drama, I just applied and easily got in. Anyway, it's like, I am not going to blame myself anymore or try to force myself too much into solutions that clearly don't work. I display neurodivergent traits and always find bureaucracy more difficult than studying. I study myself anyway. But here I am.

What is the point of grinding when you are losing so much by external standards? The harder you grind, the better university you get into, the more restricted and tired you are, the less you learn, the less time you dedicate to studying. An Asian bubble is the last thing I want. If I get a CS PhD in a mediocre university in the US that my parents and peers looked down previously, as long as I am happy, it is totally fine. I am still smart, I can figure out problems, and I am going to a luxurious industry.

I am studying math or cs all my life and I think systematically and logically, but I hate almost everything about the dominant CS culture, especially the nerdy, utilitarian image.

For neurodivergent people, one second they are find it effortless where everyone else is struggling, and then they are failing the most basic tasks like an idiot.

No Social Life and Cultural Detachment

The elephant in the room is the lack of social life for 10 years. In my middle school and university, there is a 80-90% male ratio, and zero liberal art majors. I was always denied of a social life, and I need some "liberal arts flavor" and vitality in the next stage. I largely learned about teenage life from books by Judy Blume, Ann M Martin, John Green, Sarah Dessen, or books such as "Catcher in the Rye", which does not exist in China. I read a lot from 11-17, but I haven't been reading anything in the past 4 years as I started using a smartphone at 17. I stopped buying books at Shanghai Foreign Language Bookstore and I changed to watching YouTube.

I was largely detached from the Chinese environment. From middle school I used almost exclusively English for my journals or personal writings. I share no hobbies with my roommates or classmates. I do not have Chinese social media except WeChat, I have YouTube, TikTok, and Telegram. I didn't read Chinese books, I read hundreds of English YA books. I do not play video games or board games. I like outdoor walking and aesthetics. The alienation was so strong that I could find nobody to talk to, and I felt a strong depression. Especially in university, as I complained many times already, my roommates are from provinces, never been abroad, has CET-4 English, bad hygiene, rarely showers and cleans, frequently hollering with games but extremely passive and parasocial, there is literally nothing I can talk with them as they usually wear headphones and play gun games on desktop computers for 10 hours a day and frequently out walking. I was watching YouTube or scrolling social media for 5 hours a day. By the way, I learned Russian to fluency in the past 2 years because I liked watching Russian figure skating and ballet, and now I can watch videos at 1.5x speed and read Wikipedia or news without problems. I am not boasting about anything or talking about politics. I am just making a point about the loneliness I had and difficulty of blending in with the Chinese environment.

Anyway, I didn't mention anything related to Russia in my application materials as I am highly paranoid about the situation right now except the "language fluency" on LinkedIn. I don't support the Kremlin anyway and I like Ukrainian. So I removed any mentions of Russian and changed it to Spanish, which I am learning right now anyway.

Self Learning vs What is Taught in School

I always learned by myself instead of being "taught" by someone in school. In fact, being "taught" something usually had an opposite effect for learning, such as high school or CS in USTC. On the other hand, if I am motivated to study something, I do not wait for anyone to teach me, I simply obsess with something for 10 hours a day, such as reading English books, writing, math competitions, learning Russian, web programming and hoarding. Nobody forced me to learn these, and usually learning it requires actively going against the environment everyday. The more stress I have, the less I study. I thought that PhD work is mostly studying by yourself and maintaining a good relationship with the adviser.

I already took DL and RL in Berkeley like 2 years ago, though I had a poor grade because I was half burned out. I thought like the AI/DS MS program is more or less about the same thing, except probably in a less demanding university. It is so ironical.

We quote Su shi

回首向来萧瑟处，归去、也无风雨也无晴。

(定风波 莫听穿林打叶声)

Some Other Words

A few notes about what I expect

1 Fear of professors

If I am at an American university, I would try to frequent the office hours to talk to the professors. Just talking to the professors a lot makes me fear them less. But if I get good scores, I am also going to be less afraid of professors. When I apply for a PhD, if I am going to apply for a PhD to another university, I need letters of recommendation.

2 Search for professors

The first thing I think is to know the different professors on campus and what they are doing. I need to know their labs and their research publications, and where I can do research or the thesis. If I cannot find a professor in my university, I need to look for professors in other universities and prepare to do a summer research there.

I need to actively search for an advisor who I can get along with to do research, or who is relaxed and do not put too much pressure on me.

3 Social life

An active social life is nonnegligible for my mental wellbeing. Specifically, I need to find campus clubs such as

  • Literature/languages/writing
  • Public speaking/discussion heavy clubs
  • Dancing

There is no such life inside the CS/DS departments. I will be very active and search for such student clubs. I do not like using the semi-anonymous Discord or Reddit or dating apps. I prefer physical, offline socializing. As long as I have an active offline social life with women around, dating and intimacy are secondary factors. I am open to dating but I do not have any time or excessive energy to manage someone else's emotions before I am fully secure.

4 Research

And now we get to research. I need to put far more time into doing research and publishing when I am in my Masters. I am commited so far to doing a PhD.

5 Money and future

We do not need to stress too much on these factors. As for money, I am going to rely on my parents for another 2 years, but after masters I am going to be mostly financially independent. So if everything goes right, my next 5-7 years will be like this in a campus.

I do not plan to join loud parties, drink, smoke, play video games, go on expensive travels, buy expensive gadgets, start a family, engage in addictions, or make my life unnecessarily complicated over the next few years as well. So I think this is how things will look like.

After that I want to get a US green card. I have a chance of getting it through marriage, H1B, or the EB research visa. If that does not work out, I can just study French to get Canada PR and citizenship in 3-5 years, and work in the US anyway.

6 Courses and self-study

I expect the master's minimum coursework at a third-tier university to not be very difficult and curved better. I plan to do the minimum courses required and I do not plan to bother myself with more bureaucracies. I need a 3.0 average to not get kicked out. I expect my average GPA to be 3.3 (B+) at least. To apply for the PhD, finding the right professor and getting publications is more important.

I do not plan to put a huge amount of time into self-studying about other random things unless I am comfortably publishing research and fully secure about the PhD.

The timeline looks like this.

  • April/May: Finish my undergrad thesis, get the F1 visa, while reading a lot of their research papers
  • July/August: The moment I get a diploma here, massively email every NLP profs, give them my resume and tell them my admission to MS (and obviously, no previous papers), hoping to get one or two replies.
  • September: Land in the US, the 15 months ticking clock till December 15th starts. Fight the jetlag, get an apartment, select courses, and try to help a PhD, any PhD student, do some work.
  • October - basically a whole year: Do the coursework while spending most of the time locked in a lab or library trying to publish papers.
  • November/December next year: Relax a little and prepare materials, , get 3 LORs, email and chat with new profs, apply again, prepare for new interviews, perhaps I can even get a free offer if the current prof likes me, and hopefully get into a top 50 university PhD where the new advisor isn't a sadist.

Some Other Words About USTC?

I associated USTC only with trauma, backward teachings, and students with no curiosity who don't know what Github is. In Cal they had dozens of tech clubs led by students teaching everything, in USTC they only have a CP club and a LUG. All my roommates here are quite passive and introverted and none of them are deep in research as well. You can use the 1/5 rule here in USTC: 1/5 of the students are passionate geniuses enjoying their shit, most of whom aiming directly for a US PhD, all the rest 4/5 are passive 10-hour-a-day parasocial gamers in deep depression. And somehow my roommates right now aren't in the 1/5, their graduation are also delayed and they are in a much worse depression than I am.

However, I looked into CSRankings and there are 20-30 USTC professors, publishing a lot in ML. This got me shocked. And I clicked on some professor's profiles and I think actually some of them are doing some interesting work. But I never knew these professors at all in USTC, they aren't teaching the hardware courses. I went back from the exchange program and took those hardware and systems courses with terrible professors who do no research and use heavily outdated textbook and ppts. That completely broke me and I was not thinking about going to research ever since, I was just trying to graduate to run away. But that does not mean there are no professors here, and some are highly competent, although I did hear that some labs are pure paper mills that use students like slave labors only to churn out useless papers. Anyway, somehow professors in USTC are some of the most approachable ones, and going into a lab is actually very easy here. I have around 10 professor's WeChat. This is quite unthinkable in the US. So I am going to complain as well in the US.

Objectively, Hefei and the extreme gender ratio are too bad for me to stay anyway. But I can't feel but a sense of loss, blame, and guilt somehow. There are massive resources right here that I refused to use and overlooked completely, though it wasn't intentional. It feels like literally another extreme cognitive dissonance that I am experiencing all the time. Anyway, going to the US is objectively not a bad idea, unlike going to Eastern Europe, Latin America, or Southern Europe, and my English is more comfortable than my Chinese. So while it did not change my trajectory or choices, I changed my opinion and withdrew my negative opinions about USTC as a whole.

Previous Posts

user@fedora ~/D/applying-for-masters (master)> tree
.
├── CV
│   ├── Jiamu_Chen_CV.pdf
│   ├── Jiamu_Chen_CV.tex
│   └── template.tex
├── LOR
│   ├── LOR-1
│   │   ├── LOR-1-draft.pdf
│   │   ├── LOR-1-draft.tex
│   │   ├── LOR-1.pdf
│   │   └── LOR-1.tex
│   ├── LOR-2
│   │   ├── LOR-2-draft.pdf
│   │   ├── LOR-2-draft.tex
│   │   ├── LOR-2.pdf
│   │   └── LOR-2.tex
│   ├── LOR-3
│   │   ├── LOR-3-draft.pdf
│   │   ├── LOR-3-draft.tex
│   │   ├── LOR-3.pdf
│   │   └── LOR-3.tex
│   ├── LOR-4
│   │   ├── LOR-4-draft.pdf
│   │   ├── LOR-4-draft.tex
│   │   ├── LOR-4.pdf
│   │   └── LOR-4.tex
│   └── ustc-logo.png
├── SOP
│   ├── Arizona-State-University
│   │   ├── Arizona-State-University.pdf
│   │   └── Arizona-State-University.tex
│   ├── Auburn-University
│   │   ├── Auburn-University.pdf
│   │   └── Auburn-University.tex
│   ├── Brandeis-University
│   │   ├── Brandeis-University.pdf
│   │   └── Brandeis-University.tex
│   ├── DePaul-University
│   │   ├── DePaul-University.pdf
│   │   └── DePaul-University.tex
│   ├── Emory-University
│   │   ├── Emory-University.pdf
│   │   └── Emory-University.tex
│   ├── George-Mason-University
│   │   ├── George-Mason-University.pdf
│   │   └── George-Mason-University.tex
│   ├── New-Jersey-Institute-of-Technology
│   │   ├── New-Jersey-Institute-of-Technology.pdf
│   │   └── New-Jersey-Institute-of-Technology.tex
│   ├── Northeastern-University
│   │   ├── Northeastern-University.pdf
│   │   └── Northeastern-University.tex
│   ├── SOP_template.tex
│   ├── Syracuse-University
│   │   ├── Syracuse-University.pdf
│   │   └── Syracuse-University.tex
│   ├── Tufts-University
│   │   ├── Tufts-University.pdf
│   │   └── Tufts-University.tex
│   ├── University-of-Arizona
│   │   ├── University-of-Arizona.pdf
│   │   └── University-of-Arizona.tex
│   ├── University-of-Arkansas
│   │   ├── University-of-Arkansas.pdf
│   │   └── University-of-Arkansas.tex
│   ├── University-of-Chicago
│   │   ├── University-of-Chicago.pdf
│   │   └── University-of-Chicago.tex
│   ├── University-of-Cincinnati
│   │   ├── University-of-Cincinnati.pdf
│   │   └── University-of-Cincinnati.tex
│   ├── University-of-Delaware
│   │   ├── University-of-Delaware.pdf
│   │   └── University-of-Delaware.tex
│   ├── University-of-Georgia
│   │   ├── University-of-Georgia.pdf
│   │   └── University-of-Georgia.tex
│   ├── University-of-Kentucky
│   │   ├── University-of-Kentucky.pdf
│   │   └── University-of-Kentucky.tex
│   ├── University-of-New-Hampshire
│   │   ├── University-of-New-Hampshire.pdf
│   │   └── University-of-New-Hampshire.tex
│   ├── University-of-Oklahoma
│   │   ├── University-of-Oklahoma.pdf
│   │   └── University-of-Oklahoma.tex
│   ├── University-of-Oregon
│   │   ├── University-of-Oregon.pdf
│   │   └── University-of-Oregon.tex
│   ├── University-of-Vermont
│   │   ├── University-of-Vermont.pdf
│   │   └── University-of-Vermont.tex
│   └── Washington-University-in-St-Louis
│       ├── Washington-University-in-St-Louis.pdf
│       └── Washington-University-in-St-Louis.tex
└── transcripts
    ├── Berkeley-Combined.pdf
    ├── Berkeley-Fall-and-Spring.pdf
    ├── Berkeley-Summer.pdf
    ├── english-tests
    │   ├── GRE.pdf
    │   └── TOEFL.pdf
    ├── 在读证明中英文_出国用_陈加木(PB21000002).pdf
    └── 普通成绩单英文_出国用_陈加木(PB21000002).pdf

32 directories, 72 files

Applying

I want to apply for CS/AI/DS MS for now to prepare for the transition to PhD.

  • A functional AI/DS PhD program (so if I slip during the Masters I can just stay), appears on CSRankings, and accessible with low GPAs (2.7, lower in the last 60 credits)
  • At least 50% female overall (NOT a "tech institute"), no big Asian bubble, green campus and not stressful

Right now is not the time to learn about the US geography and culture differences, campus tours, literature clubs, or humanities departments of every university. This is just procrastination. As long as the school has a good female gender ratio and a functional CS program, I will apply. I'll decide when I'm actually admitted by some. I have a list of around 40-50 schools. After crossing out those whose deadlines I missed, I have around 20-30 left. It's already March.

  • [x] Transcripts & English Tests (2 exchange, 1 USTC, TOEFL, GRE)
  • [x] CV
  • [x] Statement of Purpose (SOP)
  • [ ] Letter of Recommendation (LOR)

I applied to these schools at March 1-2 (No LORs)

SOP (Brandeis example)

I was admitted to the University of Science and Technology of China (USTC) at 16 through a difficult program for students with math competition backgrounds. I initially chose the math major before transitioning to Computer Science in my sophomore year.

In my sophomore year, I applied and went to a one-year exchange program at the University of California, Berkeley. I adapted quickly to the open-ended, project-heavy environment. Despite being only a sophomore, I took a total of 6 highly demanding upper-division courses in a year. I earned an 'A' in CS 170 (Algorithms) and CS 161 (Security), and strong grades (B+) in both CS 188 (Artificial Intelligence) and CS 182 (Deep Learning). I was deeply engaged with AI and Machine Learning, and I took a graduate-level course in Machine Learning: CS 285 (Deep Reinforcement Learning). In CS 285 (Deep Reinforcement Learning), I moved beyond theory, writing implementations for Model-Based Reinforcement Learning (MBRL), Conservative Q-Learning (CQL) for offline RL, and Soft Actor-Critic (SAC) algorithms. Despite being a little overwhelmed by the course at the time, I reviewed all the materials and did all the experiments myself again after the semester finished.

I am always interested in applying the knowledge I learned in class.

After studying web security in the Computer Security course, I discovered and reported a critical XSS and cookie-hijacking vulnerability in my home university's official educational platform, which could allow students to log in as TAs through parsing markdown into HTML without sandboxing to steal the cookie.

Inspired by the LLM fine-tuning techniques in the Deep Learning course, I tried fine-tuning LLMs including Cohere Command R and GPT-4o on my personal blog (150k words total at that time) through different approaches, such as generating a question for each sentence or paragraph, and feeding the first part of sentences as the question and the later part of a sentence as the answer. I studied how models can imitate a human blogger and how they can retain information after imitation.

I always had a strong desire to scrape and analyze online data. I scraped around 170k comments from Douban (an online forum) discussions to analyze the contents and tones of users, and I built a frontend with Next.js to search for terms efficiently in real time, as the website does not have a decent search function.

To me, computer science is far more than just a sterile subject in school, but is woven into everyday life and utilities.

My old blog was on WeChat, which, unfortunately, is very inconvenient for me: it is prone to censorship, you are unable to edit posts later, and it does not support external links. After learning full-stack development, I migrated my blog to my own website with Markdown, CI/CD from GitHub, and a Postgres database as a backend.

When I was learning Spanish, I wanted to have dual subtitles for videos. So I built an automated pipeline utilizing Cloudflare R2, AWS EC2, and Docker to scrape and process over 400 GB of diverse video content from YouTube. I optimized instances of Faster-Whisper and Helsinki NLP on cloud GPUs, generating dual subtitles and writing them back via ffmpeg. I then built a full-stack Go/Crystal and Next.js platform to host and stream the data for learning Spanish with dual subtitles. Although later, I found out that YouTube's built-in subtitles work fine, and dual subtitle display is achievable through a browser extension. As I found existing extensions inconvenient or paid, I built an open-sourced extension for dual subtitles on YouTube on my own, which now has 40 users.

I believe in the openness of computing. I spent time reading the original Git source code to understand its raw index, blob, and tree implementations. I tried different open operating systems including BSD and the mobile PostmarketOS. I experimented with self-hosting open-source software, and hosted Mastodon (social media), Gitea (git service), and Metabase (a data processing tool) on AWS, in an attempt to build a parallel self-controlled, open social media system for myself.

My traditional academic path is not without struggles. Returning from Berkeley's ecosystem to the rigidly structured, highly traditional style of my home university resulted in a period of academic struggle, delaying my graduation by a year. The experience taught me that my true motivation does not come from chasing GPAs, but from my intrinsic passion and curiosity. I spent the summer at HKUST researching the relevance of physical data and location detection. My time as a summer research intern at HKUST taught me that I need more formal research experience on top of my engineering skills.

At Brandeis University, I am drawn to the M.S. in Computer Science and the Brandeis Lab for Linguistics and Computation. I have a special linguistic background that I believe is incredibly valuable for computational linguistics research: I am a native speaker of Chinese. As I immersed myself in English in primary school and middle school, reading hundreds of novels and writing many book reviews and travel blogs on my website, I now mainly use English as my primary language for writing and living. I am actively learning Spanish. I am drawn to the intersection of NLP research and linguistics, including the work of Professor James Pustejovsky and Professor Nianwen Xue. I am looking forward to studying in the picturesque, historic Waltham city and Boston area in New England. I am ready to fully dedicate myself to graduate-level research.

I thank the admissions committee for your time and consideration.

Update Mid-March

In a span of 2 months, my mindset has drastically changed.

  • In January, I was worried about passing the digital circuit final exam and several other courses. It is a sophomore course and I was retaking it. If I failed, I would not be able to graduate this year. I did not yet cared about applying. Fortunately, through tremendous amount of effort, I got 67.
  • In February, I found out I passed all courses last semester and I only had computer organization and the thesis left. It became clear that I can soon graduate if I pass the course this semester. I began to think about applying for MS and wrote the first part of this blog. I wanted to quickly find any R1 US school to get out of this male-dominated, nerdy, provincial, and traumatic environment this fall as soon as possible. I wasn't really thinking of anything else.
  • Then I suddenly got admitted by NJIT and DePaul. And I had around 2 months to unpack the extreme pressure, anxiety, and depression. And my academic probation cleared. And suddenly I am right back on track at the previous path of pursuing a PhD in the US, just delayed by one year and probably going to a less prestigious university. Somehow this goal was abandoned for a long time. And now my goals and priorities are perfectly clear again. I dropped a lot of schools.

I got a little unhinged when I realized that everything were essentially "cleared" and I am heading out of this male-dominated, nerdy, anti-social environment in such a short time. I could only sleep for 5-6 hours a day for a week. It kind of mirrored the end of my middle school in a math-competition-heavy, male-dominated, high pressure class and I thought I was escaping as well. At that time I slept for 6-7 hours a day and I would naturally wake up at 5 or 6 am energized. Spoiler: although my high school class was gender balanced and I only stayed there for one year, it specialized in Gaokao, not in math competitions or applying to the US, and last time went catastrophically bad. Anyway, my sleeping schedule stablized back to 8 hours now.

Поховайте та вставайте
Кайдани порвіте
І вражою злою кров'ю
Волю окропіте

(Заповіт, Тарас Шевченко)

I am probably going into AI. I hate hardware and systems now, and it's hard to leverage my advantage in math there. I don't even want to do robotics because I hate hardware. I don't want to do CV yet because I don't use CV or video generation much in my daily life unlike LLMs that I am obsessed with, though I am certainly open to intersections. So for now it's NLP.

For doing a PhD, finding the right advisor/lab and publishing articles are the most important factors. I missed almost all PhD deadlines and I have no publications now. I am still applying to an MS, but now I am thinking of Masters as just a mini-PhD where I just research for 2 years. I need to go to a school either with a huge CS department or where I can find an interesting professor/lab on CSRankings to date around with before the PhD "marriage". The absolute ranking doesn't matter as much as the specific NLP lab.

Beware of those "interdisciplinary ML" professors, especially with "bio", "health", "chem", "physics", "iot", "finance", they likely have no idea what they are doing. My mom and my aunt work in healthcare/medicine. These professors are essentially publishing a lot of useless shit. I met such a professor in the HKUST summer research during the height of my depression, and my depression and sleep inversion got worse where I sometimes went down to the shore at 3-5 am and then watch the sunrise. You will think it's your problem when the professor scolds you and your self-esteem will be destroyed, or you will just drop out. And I couldn't look at that blog without getting emotional in the mix of nostalgia for beauty of Hong Kong, and deep isolation and depression at the time. From my previous blog: https://jimchen.me/a/7615fb

So like there is a grad student who had written some TCN code, and I was supposed to look at it. But like there were too many obvious mistakes, like, just in his code in almost every blocks.

Like these mistakes include: Putting 2 GB datasets inside Git and refusing LFS or S3, Using a sliding window but throwing away 99 percent of data by hardcoding the length you need to traverse to get the sliding window, Mixing the Training and Validation data, Cheating by manipulating the testing data.

Well these mistakes makes me seriously question the Integrity of HKUST, and like how crappy this school's PhD research really is? I thought as a novice everyone should be better than me, but what the hell is that?

I don't mean anything bad for HKUST, in fact, such professors are everywhere if you look on CSRankings. Institue prestige and good grades do not guarantee that someone knows their shit. In fact, the more people sprinkle themselves with empty words, usually the less passion they actually have for their field.

Because I was fighting an existential crisis of hardware classes, I became nihilistic and fatalistic. It was just like fighting Chinese, English, or Chemistry in high school. Finally the period of crisis is near its end. If you asked me 2 years ago about what I want to do, I would just say that I am overwhelmed by the courses and I know I wanted a PhD but I couldn't figure out mathematically how to do research, find suitable labs, and pass the exams in less than one year. Then obviously I failed. I wanted an MS admission to give myself some space to breathe from the year abroad and think about research then. But the courses only strangled me more. My parents thought since I could manage Cal's upper divs easily, I should have no problems passing the course here, and heavily pressured me into doing research and a quick graduation. Though that was my fault as well because I had never took any meaningful CS course in USTC except C and data structures. Then I selected 10 courses in one semester and I literally shut down.

I wanted liberal arts, but I still deeply believe that hard science is the most important thing Lacking a social life didn't kill me, but it was the Chinese bureaucratic hardware and systems courses that killed me. If I am in a provincial eastern European city or a red state city in America without a university, I would be bored to death. Languages, literature, history, geography are things I do as an entertainment or to disguise as a polymath. Because I have so much passion for knowledge, the passion naturally spills over to different cultures and beauty. But at the end of the day, I am fundamentally a math/cs person, and I realized back in high school that losing this identity is worse than losing anything else. The fact is in pop culture, a math/cs person is a nerd who doesn't actually have passion. When I talked about hooking up a vps, scraping data, or reading source code, my classmates have no interest in any of that. They view GitHub strictly utilitarian, which shocked me. Many professors are like that as well. But societal archetypes do not dictate the objective reality of who I am. The Chinese bureaucracy machine cannot erase my identity, and I plan to stay in the US for a while if I can. In fact, it's because I root myself so deeply in this identity, that I cannot compromise sometimes with the ugly reality. Humans are built for a purpose, without a purpose, or when the purpose is fundamentally not achievable, you will have a deep void and descend into nihilism. No amount of reasoning can save you from this state except for the return of a purpose.

Perhaps the main difference compared to 2 years ago, or 6 years ago when I was entering high school, is that the loneliness problem got much worse. I literally cannot communicate with my classmates in USTC. I am too aggressively westernized and we have complete different hobbies, speak in different languages, and most students here are too burned out, gaming all day in their dorms. But we are just going to the US and let's not expect everything to work out smoothly.

In short, if I can go to a suitable AI lab in the top 4 right now, I would not hesitate at all. The doubt came from choosing unsuitable environments. For example, USTC undergrad is objectively bad for me, and I would trade a mediocre US state school with it, where I can do a lot of AI research in undergrad and be in a much better position. Or for example, I would go anywhere than the fraudulent HKUST lab, or any fraudulent health labs out there. Targeting better schools and better research environment is just human nature. I thought I was a failure at Berkeley when my peers were all doing research in their sophomore years and publishing, while I was getting Bs and Cs in the upper divs. But objectively it was a success as I took 6 upper divs in CS, while the undergrad requirement in Cal was 5.

Then why aren't I nihilistic? I went to USTC, an objectively good school in China, and found it a total failure. What's the point? The point lies in the complete incompatibility of me and the Chinese culture. I found success only in math in China. In short, if I don't do math in China and go to any university to do cs, maybe except for modernized ones like Shanghai tech, I'll be equally or more miserable as well. This lesson did not teach me to be nihilistic, hate computer science, or switch to literature, it taught me to move out of China if I can. If by some chance, my parents can't afford the MS tuition, geopolitical factors, or just magically I am confined to China, I am switching back to math to prevent a total mental collapse, let alone earning money. I would probably head to the MS entrance test. I believe I can pass it if I take a gap year.

In short, for PhD

  • I am looking for a professor/lab where the "Research Interest" is quite specific, does not change drastically, and isn't filled with buzzwords. I must be able to understand the research easily.
  • I do not want to work with "paper mill" professors (who publishes just too many). Check for "tenure-track". I know NLP is a competitive field, but I can go to tier-3 R1 schools or do another field, and I certainly do not deserve to be in a state of panic or terror.
  • I am currently allgeric to "bio", "chem", "health", "interdisciplinary" (except for linguistics). I know I would drop out of such a PhD. I would rather do research in theory, applied math, security, or even pure math. In the very least, it needs to be strictly math/CS/NLP focused.

And so far, I am searching for labs or professors on CSRankings. After deciding to go to a school and graduating from USTC, I will immediately contact the relevant professors. Also I can look them up on PI rating websites and email the grad students to learn about the lab. If I can do some meaningful research during MS, I can potentially go to a better lab for PhD.

Somehow NLP is a very elitist field, like, almost all research are concentrated in top schools, unlike for example, "chem" or "bio" where everyone can pretend they are doing something. There was only one professor actively publishing in NLP in NJIT, who went away back to China. So it is very likely that I go somewhere and find zero professors in NLP, but it's fine as I can do CV or general LLMs, just no "healthcare".

Academic research is still absolutely the most important thing in the world. Failing academically, not failing socially, is the worst thing that happened to me in USTC.

I heard academic is just a very small circle in a specific field and everyone knows each other. And how LOR matters the most in the PhD application. It suddenly made me think academics, politics, corps, elite sports are all power structures and share similar features.

GPA Minimum and Direct Emailing/Calling

If the university explictly lists "3.0" as a requirement, I will send an email and call them to confirm. So I am in active email/phone contact with them. Most of these universities are still open to applications, so I am still within time.

Sometimes they write "3.0 undergrad GPA" as a requirement on their website, but when you email them and you're like, "Can I apply with an undergrad GPA lower than 3.0?" Augusta was like, "you can definitely apply", Iowa was like, "That would be up to the program". Auburn was like, "we typically requires 3.0, you should demonstrate significant strength in other areas," Idaho was like, "the grade must be 3.0, but you can be considered if you write a statement or show your experience". West Virginia was like, "We offer provisional graduate admission below 2.75 GPA", I called Syracuse and they were like, "Absolutely, we offer admissions to 2.5, 2.6 applicants." I called Kentucky and they were like, "Yes, we accept GPA below 3.0 but it must be above 2.75." and I said, "Can I still apply with GPA below 2.75?" She was like, "Um, yes, but you have to talk to the department director Simone." Then she sent me an email and be like, "I copied Simone so he is aware of your inquiries." I called Washington State and they were like, "yeah, you can definitely apply, we wait until all your materials are submitted then look into it." I called University of Oklahoma, and she was like, "yes, sometimes if you provide your background or history. We have a minimum GPA but you can reach out to the department director Jean. They do it case by case, they won't tell you whether you can or not on the phone." As it was like 4 am and I was yawning and I thought I accidentally called Oklahoma State and I called the same number again and it was quite funny, she was like, "you just called and asked the same question?"

Almost no mid-tier R1 universities I am applying to have strict cutoffs. For some schools you can never reach them by phone or email at all.

Different Flavors of Programs

I thought about these MS programs and how get into PhD

NYU Tandon, Washington, UChicago MPCS, Northeastern satellite campuses all seemingly want to put you directly into the industry. They clearly don't want you to do any research. I don't think there's any free lunch here with the brand name. ASU has literally a million students. While it is not entirely impossible to find an Assistant Professor and do some research, there will be a million students fighting for crumbs. UC schools' deadlines had mostly passed.

I looked up the grad school enrollment stats from NYTimes in 2023. Here are some schools with the highest International student percentage (the "holy crap")

  • Stevens: Total 3392, International 88% (excuse me?)
  • NJIT: Total 2475, International 81% (asked for your WeChat handle during the application process, gave a "10k" scholarship to me, which I think it gives to everyone accepted)
  • U.T. Dallas: Total 6987, International 71%
  • Northeastern: Total 13881, International 69% (I think even if you remove the Seattle, Silicon Valley, Oakland, Maine, Arlington campuses, it is still very bad even in Boston. I got an email saying "You’re invited to the Chinese student meetings in Shanghai, Beijing and Guangzhou")
  • NYU: Total 21155, International 50%

To be fair, Brandeis is also on the list with 1212, 44%, but it was known mainly as a liberal art college? Honestly don't ask me, I haven't been there, I don't know.

Here are some ideas

  • Flavor 1: (xxx Institute of Technology) NJIT, Stevens, WPI, RIT, etc. They have easy admissions and decent researchers. But I don't want to go to a tech institute now.
  • Flavor 2: (a private university) Brandeis, Tufts, Syracuse, Emory, WashU, etc
  • Flavor 3: (a medium R1 university) Vermont, New Hampshire, Wayne, Iowa, Oklahoma, Arkansas, etc

Then I recalled my experience in Cal, and why I felt like a terrible failure that year and how I did no research. There were like a million undergrads and exchange students from Tsinghua, Peking, already with publications, all trying to get into research. "URAP" (undergrad research) is "open" to exchange students, but it is a joke for you to get in. I had zero CS experience at all and just changed my major.

In the second semester, I still didn't get into research, and several other aggressive students arrived. There was literally a number 1 kid in GPA from USTC there (now at Stanford), and several other extremely competitive people from USTC, fighting to get into research. I didn't know if they succeeded right in Cal as I largely ignored the group to protect my mental health. Then there was the DL course, and they were always in the same room in Soda Hall. I began hating Soda Hall. Then there a research event with like 10 research opportunities for top NLP/AI, but the hall had a thousand students in it. It was a sad story. My favorite places were the West Berkeley shops, Emeryville, and the Durant Avenue 1-2 am snacks.

Anyway, let's not be self-defeating and it isn't 100% that I can't get into research if I hustle, but this is just objectively a very uncomfortable position to be in. You don't want a terminal MS program there. Cal is genuinely a good place. The whole campus is alive with dozens of different clubs (though I got into none, they required two rounds of interviews) and people having fun with tech, while USTC has zero culture like that. Cal students are actually far more competent and aggressive in CS, while their undergrad courses are much more lenient. I spent a lot of time at Moffitt and the "main stacks" cold-war bunker. So yeah, MS in a university with a million CS undergrads/exchange/MS students or a terminal program is a very bad idea for my PhD dream.

Didn't Apply (Got first admission)

I made a huge list first, but after getting the first admission I dropped a lot of schools in the midwest or with tiny CS departments.

Deadline Missed

LORs

If they agree to write LOR for me, I will thank them and buy them some gifts. These are the teachers I will be asking for LORs.

  • [x] Bei Hua (Professor, Thesis advisor)
  • [x] Lei Gong (Associate Professor, Head Teacher)
  • [x] Shuai Shao (Professor, Previous Head Teacher)
  • [x] Baizong Wang (Instructor, got 99 in Intro to C)

I will send the LORs to them at once in 2 phases. There are actually a lot of fear in contacting with the professors here directly. I failed too many courses and I rarely talked to them, so I need to frequent office hours later.

Final Applying Results (23)

March 1

  • New Jersey Institute of Technology (No LOR): March 12 🟢 Admit (AI)
  • DePaul University (No LOR): March 17 🟢 Admit (CS)

March 2

  • George Mason University (No LOR): March 27 🟢 Admit (AI)

March 17

  • Arizona State University (No LOR): April 9 🟢 Admit (AI)

March 19

  • Wayne State University (No LOR, No SOP): March 23 🟢 Admit (AI)

March 24

March 27

March 28

  • University of Arizona: April 3 Dept Admit April 16 🔴 Grad School GPA Reject (CS)

March 30

March 31

  • Brandeis University: April 13 🟢 Admit (CS)

Choosing a School

You don't roll a dice or rely on gut feeling. You look up preferred labs on CSrankings, and email a professor, be like, "Do your lab have openings for incoming MS students? I am admitted to the CS/AI MS program, and I already took DL, RL, Algo, Linear Algebra." If they agree, then go.

PhD Interview

Professor Xintao Wu from University of Arkansas sent me an email on April 2nd saying he would like to have a 45-minute interview with me as I mentioned wanting to join "SAIL" in the SOP. He was a professor at UNC Charlotte till 2014. I looked up the professor and he is also an alumni of USTC and graduated in 1994. He has the "Acxiom Endowed Chair" (which means funding?) and is the top publisher in Arkansas on CSRankings (basically because that's where I looked up professors and put them in my SOP). I thought usually you don't get interviews with professors when applying to masters as there are thousands or more applicants and professors are busy. Even if you do, I think it is very rare for a professor to reach out first when I didn't even email the professor beforehand. And somehow I went on the website of Arkansas and it said students "must have a 3.0 GPA from bachelors", and the "SAIL" lab also listed "3.0 grade" in the requirements. So I went to read some papers for the interview.

So I spoke with Professor Wu for an hour on April 3rd, and we talked a lot, but mostly in Chinese. He asked about my academic background and my failures in school, and I explained that I was in a bad state after I went back to USTC and I said I am considering AI. He asked if I would be graduating, I said yes and I said the transcripts showed me retaking and passing courses. He expressed understanding and he said he is fine with that as I am from USTC. He asked how I found out about the programs in Berkeley and HKUST, I said I applied and mostly paid for myself in the Berkeley exchange program with some reimbursements, while the HKUST program was a school program, but you still need to apply. He asked about how I found Arkansas, I said CSRankings. I asked about 2 papers. He asked me about the projects I wrote in my SOP and my grad thesis and I told him about it. He asked me if I can code, and I told him about my projects, and I said I can implement ideas mostly without much problems. I told him how I implemented a transformer and I could put different book texts into it and generate something with the author's style. So he is mostly doing LLM privacy, and I asked about the "unlearning" process. It was about "attention masking" or something. I asked if it worked like, "John Smith lives in this street" and then hiding it. The professor said it doesn't work like this, but work for a specific record classification for copyrighted labels. He said that is partially the work of another vision transformer paper about hiding the face or something. I said I was curious if he tried to jailbreak it in different languages (such as "Де живе Гаpрі ПотTэр?") and he said they didn't really try it. Then I asked about another paper, which is from a student who worked in Walmart. Though, he discouraged "free exploration" and reading "systems code". He told me about the offers like Amazon or Google, or tenure tracks his former PhD students got. A lot of those students are Chinese and he said some went back to China. He said that his students are all well above the "average level" in Arkansas. (fair enough, he and another Postdoc student in the group are the top publishers on CSRankings in Arkansas by a huge margin) He said he have 1 Postdoc and 4 PhDs currently, 2 of his PhDs are graduating, and he is hiring 2 more PhDs this year. He already had one, and he is considering me. He asked me about my plans and he said he can tell the committee and transfer me directly to PhD. He said that MS in America is usually 2 years, and you still have to do a 4-5 years PhD after that, if you do PhD here you can do a PostDoc in a better school. I asked about whether his PhD students are being "pushed", he said the Chinese students work really hard, but if you don't want to work hard and just want a job, he will try to help you succeed as well. He said he have 3 million dollars endowment and he can use the interest for funding. Then we talked about curiosity and exploration. I said I have passion for computer science. He said you must publish in a PhD. I said I should be more modest. So basically he is suggesting me be more utilitarian. I said it and I was like, "You mean like being instrumental?" He smiled a little and he was like, "You are young and you can afford a few years exploring, you certainly can have curiosity, but they may end in failure." He suggested that you have to work much harder than the local Americans for success. Anyway, I said I was really thankful how a professor took an hour to talk to me and I said I thought professors are busy. He said if I got better offers, I can certainly take it. Then the meeting ended.

I received an email, where Professor Wu emailed the Department Head Roy McCann that he would like to have me in his group with "potential RA support", and Roy McCann (Head for Academics & Graduate Coordinator) told me to email the Graduate Admissions Office and provide the GRE score. I immediately confirmed my intention to join the lab. Then Jennifer Sandridge (Director of Graduate & International Admissions) completed the academic level change of my application to the PhD program. Then I went to the application portal again and uploaded a few empty PDFs for some requirements, and it went into the security clearance process.

Let's be completely honest, there doesn't seem to be any "hacker culture" or deep passion for CS in Arkansas (reading soure code does not register for fun to them), the culture does seem utilitarian, and his English was heavily accented (he switched to Chinese in the first minute). But on the plus side, he is already tenured, he has papers in ACL/EMNLP/AAAI in the last 2 years, I get good funding, I literally get a direct PhD, it is a gender-balanced university, and I think he is relatively lenient and cool about graduation. And for my specific profile at this moment, it is far more than what I believe I could ask for. If I go to a regular MS program, I have the huge burden of fighting to get research with just any NLP professor in a Capitalist program, getting a good GPA in the required courses (it is not "trivial"), and finally I might still land at a utilitarian, high-pressure, possibly a tenure-track, sadist professor, and start the PhD only by then. If I go straight to a PhD, I can quickly pivot my focus to reading and publishing papers.

Basically it is futile to argue right now. I sent an email to Karuna (a student graduating this year from the lab) and Min Hao (a student who graduated last year, now in Google) scheduled a video chat. From the professor's perspective, I am an "unknown variable" who is perhaps good at math but very arrogant, uncooperative, lacking in self-awareness, and who failed a lot of basic courses. The last thing I want is to alienate or irritate my potential advisor when I have no leverage. The most important thing right now is publications. If I can't publish a paper, the burden is fully on me and I would fail the PhD anyway. I must be cooperative for now. As long as I am actually publishing something, I think I can have some freedom and the rest of the time is mine.

Lab Culture (Per accounts of the students I spoke to)

  • Work-life balance: Very pushy or hands-off? Work hours in the lab?
    • Weekly group meetings and 1-to-1 meetings, people are in the lab for several hours a day (half a day maybe?) but not 996. If you aren't meeting your potential, Dr. Wu will tell you to work harder.There are moments where he will tell you your idea is wrong but you really want to do it.
    • It depends. Expected to come from Monday to Friday. On average 8 hours, but nobody cares if you are missing. Professor has own office. 1 meeting with you per week, 1 meeting with the whole lab.
  • Lab culture: Collaborative or people work independently? Do they speak Mandarin?
    • You can choose to collaborate with other people. You can also choose to collaborate with other faculties. (actually, Dr. Wu and another Postdoc in the university are both from USTC, ok) But every paper usually has a main "first-author". Actually there is only one part-time PhD student from China now, the rest 4 from India, Vietnam, Nepal, and South America, 2 are graduating, so they mostly speak English (but somehow the professor interviewed me in Mandarin). The professor had a lot of former Chinese students though (and from his time in UNC Charlotte).
    • We have collaborations. There are other Chinese professors Dr. Wu is working with.
  • Funding: How much a month? TA vs RA? GPU?
    • Stipend is 2000 dollars a month. Says you have to find a roommate. (really?) Has a scholarship but she thought it passed for this year. 4 years RA, 1 year TA. Has some funding from the chair. Has some NSF grants, they are ending. He is writing new grants. Has university wide V100, A100, and some H100s. Didn't tell me about the details though.
    • Not sure, ask Dr. Wu. Dr. Wu allow you to do internships, but you can also stay in the lab to do more research. Students take internships in the summer and Dr. Wu allow that.
  • Graduation
    • Professor Wu doesn't keep you around like a slave labor for 8 years. If he thinks you are failing, you will "master-out", as seen on the "Alumni – MS Graduate" at the bottom of the lab page. If you are doing a PhD, you usually graduate in 5 years, or 4 years if you had a Master's.
    • Once he feel that you are ready to graduate with 4-5 papers, you can do a dissertation, and he will not keep you for long. He is fair to students. In the summer you can stay in the lab, or you can do an internship.

Admission Delay

MS send offers really quick, sometimes within a week. However, incoming PhDs need to go through a security clearance process as I am from China. So I paid the deposit for WashU and Brandeis.

If I must go to an MS, I will probably go to WashU or Brandeis. WashU sent me an email about how their students are receiving PhD offers. They posted the students' MS thesis on their website. Brandeis sent me an email with a "faculty spotlight" PDF and an hour long video for CS students where each professors talked for a while. I called Brandeis and chatted with someone named Anna. I asked her how many students they have for MS and PhD, and whether there are over a thousand MS students. She said there was only 10-20 admitted masters in CS each year. I was so surprised and I said that is a huge relief, then I realized that I was one of the less than 20 admitted students. So there are indeed opportunities for research, and it looks appealing to me.

Chat #2 With Professor Wu

Mainly about research. If I get the official admission, I will likely go. Thanks professor for the time and for advocating to the security committee for me.

Prof: Due to ... (blamed the government), Chinese students need a long security check these days.

  • What direction will I be working on? Since I will have RA funding, which specific grant or project that will be funding me?

AI for science (I don't want to do it!), also has endowned chair, but I can also do trustworthy learning (professor claimed is "crowded").

I be like, "I didn't study one course in Chemistry and Biology and I know I am bad at it." Professor be like, "back in our times in USTC we need some Chemistry and Biology." Professor be like, "fine, I am just telling you the situation, I am not forcing you down a path you don't like. I am just saying it is more crowded there. You cannot compete with big labs in fundamental research anyway."

  • GPU resources

8x H200, A100 (several years ago), shared HPC

  • Do I have flexibility for my research? Or is the research strictly defined by the lab's projects?

Due to ... (blamed the government again), NSF/grants are harder, Professor has much less funding than a few years ago. As Professor bought 8x 200 GPUs with some 300k dollars, the money is more tight right now. "AI for science" (claimed is the new direction) can pay get grants. But you can still choose your own direction more or less.

Professor said you will not have time to collaborate with other professors or do flexible things once you get bogged down in a project or research, where you have to produce results in several months and compete with other researchers.

  • Do you expect students to publish in their first year, or mostly focused on coursework and paper review?

Directly into research. Professor says you can graduate in 4 years if everything goes well. There are almost no student who did "TA" except Karuna, almost everyone is always an "RA". Most people listed on the lab page graduate in 4-5 years. We got a student from Bangladesh and a student from China coming this year. You might be collaborating with them.

About Finance

As I do detailed finance accounting every month for any amounts greater than 50 RMB, I have to figure out "who pays for what and how much".

While my parents are always comfortable upper middle class in Shanghai, they are extremely stingy with money, relative to their net worth. They like to use money as leverage over me, frequently using gaslighting and emotional guilt to deny me of making my own choices. I couldn't wait to be financially independent. Relying on my parents would stop me from having true independence and adulthood.

2-year MS cost in the US (out of state)

I do not live lavishly, nor am I suffocating on a budget. In my exchange year, I lived in a separate bedroom or studio and eat around 2 meals a day in restaurants, cafeteria, or the supermarket. I spent at least 80k USD during the exchange year in Berkeley (I took 10 courses total which is over 30k, Berkeley extension fee is 10k, rent is around 20k, and I went to Alaska for around 4k USD, I didn't do detailed accounting back then) and I got around 8k USD (55k RMB) reimbursement from USTC.

I calculated the fee of an MS in the US and it was anywhere from 80k USD to 160k USD, let's say, 500k RMB to 1.2M RMB. The credit hours are anywhere from 1k USD to 2.5k USD. Some universities have a "flat-rate" fee (33k per semester in WashU for full-time students). So the course fees are anywhere from 30k to 100k. In short, it depends on where you are going (New York, New England, California, or a deep red state like Oklahoma) and it depends on the specific school. But you'd be paying 15k-40k a year to the school, and spending 40k-80k USD a year total.

Capitalism likes this business, so admission is usually incredibly easy for MS. I got admitted to so many schools. A single school like Northeastern can graduate around 2k CS MS every year. Other schools such as UT Dallas, NYU, ASU, NJIT takes 500-1500 CS MS students I think, mostly International.

How does the university gets money and who funds a PhD? (from their perspective)

Professors get money from NSF, DARPA, or other grants, or money from industry connections. However, university will tax part of that money for itself (let's say, 50%). A prestigious professor receives an endowed chair title and can get money from the interest rate (usually without taxes).

A CS PhD costs a professor 60k-120k USD a year (depending on whether it's a deep red state or a rich blue state). That's around 300k-600k USD in 5 years. This covers the student's stipend, tuition (you take courses mostly for the first 2 years, but then I learned about "phantom tuition") and healthcare insurances. That does not include research budgets such as AWS GPU credits. So the popular labs such as BAIR or Sky in Berkeley are operating like mid-sized tech companies.

Capitalism leaves rich majors (such as CS) to the "market demand", so admission is usually incredibly difficult for CS/AI. There are around 2k total CS PhD graduates a year. The very same schools (Northeastern, USC, ASU) has around 40-80 CS PhD students a year.

  • TA vs RA: I think a CS PhD student spends most of the time as an RA, and maybe a few semesters as a TA. This is not like undergrad TAs who work on an hourly rate, though I was never an undergrad TA anyway. Poor majors (such as liberal arts) spend most of the time as a TA on a minimal salary.
    • If you are an RA, the professor pays.
    • If you are a TA, the university pays.
    • If you have a serious medical problem, the insurance company pays.
  • However, PhD students usually cannot work more than 20 hours legally a week on campus. A typical TA or RA job is 20 hours. (but obviously you do research for far more than that) So it means you work either job. You always have to do research, while you have to teach if you are a TA.

Tuition and campus life

I think a PhD student takes courses in the first two years like an MS student.

  • What if I take a "literature" class? Who pays for it? I don't know. Maybe you have a tuition waiver up to a certain amount and then you have to pay. Maybe they have a flat-rate tuition. Maybe the advisor will stop you from taking any "philosophy" courses. But I probably won't be enrolling in classes because I hate listening to a lecture without discussions and being graded.
  • Can I audit? Depends on if the specific professor or instructor is fine with it.
  • Can I go to a literature or writing club meeting for 2 hours a week? I think I can if my research is going well and I am publishing papers. But I probably won't be "leading" a club.
  • Greek life? No.

Normalized salary

Let's calculate the "normalized salary". I think the tuition counts as your salary if you are taking classes (because otherwise you need to pay to go to classes), while phantom tuitions and healthcare shouldn't count (because you don't need to pay for it with a take-home salary). What is your actual salary compared to the average personal income in the US?

So let's say, in a deep red state

  • 45k USD for the first 2 years
  • 25k USD for the next 3 years

That's around 160k USD total, so around 30k-35k USD per year. That's just slightly below the average wage in deep red states (around 40k USD per year). You can also work internships in the summer if things are going well. I don't know about the exact number or if I would do an internship, but I think it is more or less around the average wage and basically comfortable.




Loading comments...