Topic

Wishing everyone a happy Mid-Autumn Festival & National Day holiday~

Since training large models takes too long—plus spending a ton of money and having a mental breakdown over it—I’m just going to ramble and chat a bit.

Today’s topics are, on one hand, how to differentiate yourself in the workplace, and on the other, whether to look for a job or go to graduate school.

How to Differentiate Yourself in the Workplace

Company A, where I work, just finished its performance reviews. In truth, everyone’s work is quite similar: engineering folks either refactor, iterate, or stack features with CRUD operations. Leaving aside uncontrollable factors like promotions and interpersonal relationships, how does the boss distribute different performance ratings among team members?

Through observing and practicing performance reviews with high-performing colleagues, my suggestions for creating differentiation are:

  • [Scarce Skills] In an engineering team, study algorithms & data; in an algorithm team, study engineering & data; in a data team, study engineering & algorithms. In short, analyze the scarce skills within your team to avoid direct competition and easily reach the top position in the team.
  • [Efficiency Tools] Observe pain points where efficiency can be improved, develop automation tools, promote them, and continuously maintain them. Over a few years, you will build a personal brand and enjoy the power of compounding.
  • [Seize & Monopolize High-Yield Projects] Treat yourself as a company to run. How does a company grow? The basics are product supply and customer acquisition; advanced steps are retention, conversion, and branding. Good product supply and conversion correspond to high-yield projects, while strong customer acquisition, retention, and branding correspond to attempting a monopoly. Make it so whenever the company or team mentions a specific business domain, they instinctively know to come to you. If your performance summary presents impressive data, charts, and evidence of sustainable development, then you’ve done quite well in this area.

Job Hunting vs. Going to Graduate School

Recently, a relative of my college roommate who slept in the same room as me for four years (sounds a bit weird) reached out to me. They said: they aren’t confident about the postgraduate entrance exam, want to retry for a second year to get into their dream university for graduate school while improving their abilities; they haven’t learned much technical skill during three years of college, so campus recruitment might be a long shot; they want to become a university teacher and pursue a PhD.

My response was also very straightforward. First, let’s set aside whether to pursue a PhD or not for now; the key question is whether to take the postgraduate exam or enter the workforce. So, what is their ultimate goal? Unless your family is sitting on a mountain of wealth, it’s impossible to stay a student forever, and one might instead choose graduate school just to follow the crowd or evade competition. Therefore, before agonizing over such questions, ask yourself: what is your ultimate goal? Luckily, they were very honest: the ultimate goal is to make big money working after graduation. That makes things relatively simple; thus, the career goal is simplified to meeting the passing threshold for big tech company interviews.

Next is the postgraduate exam. Since they have a top-tier university complex, they plan to study for another year. I can understand that—after all, I also wanted to go to a 985 or 211 university to try to rewrite my destiny, and I once suffered a total mental breakdown before the postgraduate exam (the kind of breakdown where opening Zhihu shows everyone flexing, while looking at actual past exam questions leaves you completely baffled). Therefore, my advice is to go all out preparing for this December’s exam, don’t overthink anything else, and after taking the exam, quickly upgrade your skills and try to get referrals for spring campus recruitment. I recently watched the movie The Bucket List; unfulfilled wishes are the most regrettable, so no matter what, chase your dream first and give yourself an answer. Furthermore, if job hunting succeeds during spring recruitment, you won’t have to gamble on uncontrollable factors like whether graduate school can improve your abilities, whether the overall environment is friendly, or whether the time and opportunity costs can be recovered.

After spending days of time and money training large models, along with quiet reflection after my mental breakdown, I realized that the first-year coursework in graduate school only teaches mathematics, English, machine learning, plus some basic knowledge of CV and NLP; the rest relies on self-study or lab teamwork. To put it bluntly, if you blindly go to graduate school without knowing what you want or what your ultimate goal is, you might not even want to, know how to, or be able to afford to run Transformers or large models, and your tolerance for error in life after graduation will become even lower.