Holy flip, I haven’t updated this blog in a long time. Anyhow, yes, I am still alive. Since I last updated this blog, I actually went to Israel for a ten-day birthright trip and met some wonderful people there, and learned some stuff as well.
I also found a job as a Research Engineer at a small Silicon Valley algorithmic asset manager. Currently, I’m in training, because there was some stuff I didn’t know about R, and my boss is absolutely adamant about great programming practice. Yep, I’m still kicking myself for not taking more computer science courses, and not being as serious about comp sci as I should have been (I thought I was going to become an actuary for 3 out of 4 years…worst decision ever!).
And speaking of computer science educations, several weeks ago, I was at a Udacity worldwide meetup, where I was hoping to meet the man that founded the company, Dr. Sebadasstian Thrun (errr, Sebastian…I still call him Sebadasstian, because it’s absolutely badass to have invented the self-driving car, and started what I think is the coolest of the 3 MOOC companies (edX, Coursera, Udacity)). He was tied up in a meeting with Eric Schmidt, but I did get to meet the COO, Vish.
And of course, Udacity was looking for feedback, and while I do have some small criticisms (please bring back the old vertical checkmark progress method, rather than these tabs across the top that are just that much extra confusion), my main feedback for my most favorite startup in the world was namely…a wishlist–simply because I’d like to see Udacity rock house and revolutionize the world in terms of education.
So, let’s get to my main wish:
Vertical silos leading from no knowledge to jobs at top Silicon Valley firms (as well as algorithmic trading companies =P). Basically, the way I envision it is this:
Employers state that they simply can’t find enough applicants skilled in computer science. Colleges charge an arm and leg for results that aren’t stellar for most people. And applying to jobs, unless you come from the elite of the elite universities with a nosebleed top-tier GPA, is often a crapshoot. In fact, I can generally envision the algorithms (here’s some R pseudocode…probably not the best language to use for natural language processing but that’s beside the point >_>)
topSchools <- students[students$university %in% c(“Harvard”,”MIT”,”Massachusetts Institute of Technology”, “California Institute of Technology”, “Stanford University”, etc…),]
engineeringMajors <- topSchools[grep(“engineering”, topSchools$major, case=FALSE),]
otherSTEMmajors <- topSchools[tolower(topSchools$major) %in% c(“math”,”mathematics”, “operations research”, “computer science”, “statistics”, etc…),]
acceptedMajors <- rbind(engineeringMajors, otherSTEMmajors)
acceptableGPAs <- acceptedMajors[acceptedMajors$GPA >= 3.5,]
citizens <- acceptableGPAs[which(acceptableGPAs$citizen==TRUE),]
(Basically what’s happening here is just a lot of subsetting)
At the moment, there are a lot of inefficiencies in the recruiting process, mostly stemming around trying to gauge a candidate’s knowledge in just a single digit amount of hours grilling the candidate after looking at one piece of paper. And while Github may help with that, for one, I’d think the lion’s share of a person’s time that programs for a living is spent writing code that either A) isn’t flashy (i.e. if I’m a data scientist, most code I write isn’t going to do something profound. It’s going to be selecting this data or that, subsetting this (as demonstrated by the pseudocode above), doing a basic operation on that, etc…) B) locked away under NDA or C) a “for-fun” little piece of hackery.
Now, when I called Udacity to get some career help, I was essentially told “github, excellent kaggle results, or bust”. Namely, I felt dismayed that although I had passed Dr. Thrun’s self-driving car course with distinction, I had completed Dr. Norvig’s (research director at Google!), and completed the “building a search engine” course with all gold star questions answered (most correct, I believe? Obviously the base class was “just” CS 101, but the gold star questions included a bit of NLP–in all fairness, for one of the questions, I actually googled up somebody else’s implementation of Levenshtein Edit Distance in Python =P ), and yet, that still didn’t matter one iota–to the person responsible for Udacity’s career services, no less. I suppose this is a fair point, as Udacity at the time was (and still is) a very young company.
However, as Dr. Thrun alluded to, he wants to make Udacity’s mission that of helping people advance their careers in technology. In fact, I really like this article:
However, I seriously disagree with Bill Gates in that anyone motivated enough to learn through Udacity would do equally well learning through a book. Nuh-uh. What I personally like (or liked, I think their format’s changed to something hideous lately, with the tabs across the top), about Udacity is the way that Udacity can teach to someone like me, who grew up on games, lights, and sounds, and thus, doesn’t have the greatest of attention spans, or ability to simply read a chapter for several hours and digest everything.
See, while most reference books may be good for just that–reference, in my opinion, most of them aren’t all that great for teaching. They have no exercises, or simply present information in a deluge. I think one of Udacity’s greatest strengths is that due to the way it’s designed from the ground up to have people do programming online (unlike Coursera), Udacity can teach and assess for learning in small, bite-sized chunks that actually have students coding, rather than simply filling in a fill-in-the-blank, or multiple-choice question every 5 minutes ala Coursera. Granted, the simple online programming environment may not be enough for more involved coursework, but that road can be crossed as well.
In any case, the fact that there’s a voice, colored text, a moving hand, and small, bite-sized assessments, along with the checkmarks that show progress for the unit is a very innovative, dare I say gamified (hello, Kevin Werbach!) approach to teaching, that in my opinion, makes Udacity’s format FAR superior than reading a book, whether paper or online. In fact, it’s also far superior to any 90-minute lecture as well.
Now in my opinion, it is wishful thinking that anyone with the drive and motivation to make it through MOOCs will be an easily-employable candidate that commands a salary respectable enough to really live well in Silicon Valley. Simply, there are (and I know from personal experience) a whole host of other factors that go into getting a job. For instance, choking at interviews.
Some people just are not great interviewers. They may talk too much, they may get nervous, they might choke up on a question they’d easily be able to answer if they were allowed to do a quick google search, and so on. In fact, the reason that I got my first-ever job was that the quant that hired me, instead of putting me under the gun, sent me several scripts just to read, and gave me a week of a deadline. In other words, no pressure whatsoever. I found the process so entertaining that I literally asked for more work. Unbeknownst to me, however, when I arrived for my in-person interview, the offer was already in place, so long as I didn’t piss anyone off in conversations.
In my opinion, this should be the ideal way to do recruitment. That is, find a way to answer the question of “is this person qualified to do the job” without putting it in the context of “can this person answer a bunch of random technical questions while under the eyes of people who will determine his or her standing in life”? After all, if a person is sending out resumes, they’re not doing so for laughs and giggles. They’re doing so because they need either a job, or a new job. And if they do, odds are, they’re probably under some amount of stress to actually find that job.
Also, there’s the issue of networking. Although the old adage is “it’s not what you know, it’s who you know”, in my opinion, it’s both now. Not only do you have to know something, you also have to know someone that can refer you. Odds are, you don’t need to have a cutting-edge elite university education to do statistical analysis at Google. But why should Google risk hiring from a non-elite university? The solution? Networking, referrals, etc…, as opposed to ability.
And on the part of the recruiter/employer, as I stated before, the problem they have is that they have to be extremely choosy simply due to time constraints. If you’re a recruiter at Google, you’re dealing with thousands of resumes, and for each candidate to go through an interview process of several hours, as well as the time put into scheduling and preparations and so on, if you have, say, 200 candidates getting in-person interviews, and each candidate requires 10 hours of man (or woman) hours, that’s one year of human-hours gone, right there. So recruiters/employers are fully justified in doing business the way they do (and why referrals work the way they do).
So, my question is: why doesn’t Udacity make it its mission to replace not necessarily the university–but the inefficient recruiting process (of which a university degree is but one aspect)?
Here’s the logic: if a recruiter judges someone on the quality of the undergraduate college they got into, well, the college someone got into is predicated on their high school performance, up to junior year. In other words, ages 14-17. Then, by the time those people apply for entry-level jobs, their GPA is composed of their performance from freshman to junior year, all equally weighted (i.e. if you had a couple of terrible semesters early on, like your freshman spring and sophomore fall as in my case, tough luck!), and from then on, someone’s employment history is most likely comprised of the fact that if they started off as a second-tier candidate, they’re going to be a second-tier candidate for quite some time, since a candidate’s current state is directly dependent on his or her previous state.
In other words, there’s a huge amount of latency on that resume. In high-frequency trading, millions of dollars are spent on technology to shave off microseconds in execution. In gaming, latency beyond a fraction of a second is unacceptable. But in recruiting? We’re talking months, or even years of latency.
What Udacity is in perfect position to do, in my opinion, is to have capstone classes, that, if passed to a certain acceptable degree, rather than just hand the student a shiny PDF saying “you passed our course with highest distinction!”, instead say “you completed our course with highest distinction! Here are special, one-time, Udacity-generated links to employers, that when you apply to them, they’ll know you applied because you have demonstrated both the self-motivation/drive, and the technical aptitude, to be a productive employee of this company.”
And from such capstone courses, Udacity can build down, right to the very basics, such that given enough drive and motivation (and perhaps talent/intuition), anyone who simply is willing to put in the time, can fast-track their way from “this is cool” to “I do this for a living”.
And rather than having to roll the dice on candidates, and simply look at titles of courses on transcripts, employers will know for a fact that Udacity itself vetted the candidates, and will know exactly the material that was covered in these courses, and so, can hire with a lot more accuracy.
And as for Udacity? Well, rather than try and get fees from candidates for “authenticated certificates”, or teaming up with universities to try and collect royalties from referring candidates for online masters’ programs, if it can produce these employment silos, and be both teacher and recruiter, well, if a Silicon Valley company hires a candidate for $100,000, and Udacity collects 10%, that seems a win for the company, a win for the student (now employee), and a win for the company that hired him.
As for my own personal silo wishlist, it’s quite simple:
Two of them–data science/statistics/machine learning/data analysis
Trading (as in, the quantitative, computer-and-rules-based Renaissance Technologies/D.E. Shaw kind).
(No, it’s not just about the money–I think that trading systems are simply cool–just the idea of risk vs. reward, controlling position sizes, the trading signals, etc… are awesome).
So yeah. Odds are, this post was too wordy. Which is why I suck at interviews =P