Conversation with an ‘Immediate Junior’
6th January 2021
I recently noticed that most of the job openings in industry are biased towards people with 5+ years of prior experience. I decided to boycott recruiters by ignoring direct messages and emails from recruiters. I decided to help job seekers by committing some of my personal time to jointly work on projects, presentations, etc. I’m convinced this is the best I can do because I’m not a job creator. However, I did one mistake - I deactivated by LinkedIn public profile for a while to avoid recruiters, which also blocked 2nd, 3rd, … contacts from messaging me, but I corrected it recently by making my public profile 100% visible to everyone (and continuing to ignore recruiter messages / emails).
Anyway, striking a meaningful conversation with me can be hard (this should be unsurprising). In the first few interactions I try to separate people with wants/favors from people with needs (just to prioritize). I have been approached by people with specific needs like “I’m seeking a job as a Machine Learning Engineer, but I’m not good at programming” - I’m 100% open to take this conversation forward because the individual has invested significant amount of time to analyze the industry, to identify his/her skills and to match skills with job requirements. However, this write-up is NOT about one such person.
I recently proved that I can return a meaningless conversation by wasting someone’s time. The conversation (not exact, but the message is accurate; ‘X’ denotes the other person):
X: “Hi, I was your immediate junior at the F-school” (reminding me of the F-school is probably not the best way to start a conversation) Me: “Hi! Ok, I don’t think we met” (I guess I know the person by name, but the statement is 100% true) X: “I’m reaching out for some guidance. I want to switch to ML field. I remember you taught stats and pursued a career in ML right after college” (my job offer was a secret; I’m 100% sure this statement based on a quick look through my LinkedIn profile) Me: “Yes, but I’m not sure if I can be helpful. I’m very theory and tech oriented. I’m sure there’s enough room for people like you (not exactly what I said) in ML, but that’s not my focus area” X: “Ok, I recently took courses in ML models (regression, <insert any fancy ML algo name that you wish, otherwise just use ‘Random Forest’>”, etc.), but I’m having a hard time applying it in real life. Please suggest courses on data processing/preparation and give me ideas on steps to take before jumping into model applications” Me: “There are several practical courses in several platforms. I’m not sure about their differentiating value because everyone with a bachelor’s degree also works on those data sets. Can you be more specific about your expectations?” X: <No reply, probably not going to receive one>
Takeaway
The most likely progression would have been one of the following:
- This is the most probable outcome of a conversation with a normal F-school grad who is only after money. F-school grads have limited skills and no goals, no intent to learn, …
- X has a data set (proprietary) and wanted free advice (and a working solution) to prove to his/her manager that he/she is capable of doing ML. This is illegal! In the past I was guilty of the same mistake once; I felt sore for being the company’s bitch and apologized to the person who provided consultation, reassuring him/her that I will not consult him/her if the previous consultation fee was not paid (which turned out to be the case).
- X wants to start up in ML and wanted free consultation (and a working solution) to pitch to clients/investors. This is illegal!
At work some of my reportees and peers had similar problems - they were trained to import sklearn, but they could not independently formulate problems and work on real life data sets. Some of them were sincere; they were willing to accept their shortcomings when mentioned in an objective manner. In some cases we learnt things together and came up with simple solutions that surprised technically sound clients; in some cases I personally invested time on the individual to identify glaring gaps in logic (not syntax) that ‘coaching institutes’ will not dare to highlight.
There’s a bubble - a growing number of people in ML who know only to import sklearn, pandas, numpy, tensorflow, matplotlib.pyplot as plt. Very few of them become good experimentalists. Others are old school analysts who carry ‘scientist’ or ‘engineer’ titles (sad truth). People with F-school degrees with these traits belong to a much more inflated bubble.
Conclusion
With limited free time in my hands, the best I can do is to work with genuine ML enthusiasts who are willing to work on first principles. Relying on my personal path to ML can be wrong - even I don’t do it; I try to reinvent the wheel several times. Libraries like caret, sklearn, etc. may be gone tomorrow; tools like Hadoop may not be relevant tomorrow; but the first principles of statistics, machine learning and computer science will not change too frequently (humbly acknowledging the fact that new radical principles will produce better results in the future).