2/ Before starting, I should point out that I hate such automated keyword filtering tools. This is precisely why I co-founded @ReliScore with @aparanjape: so that companies can skip such automated filtering and directly move to testing for *relevant* skills.
But...
3/ Most people on reddit/hackernews/twitter who complain about or laugh about these problems with the recruiting process don't understand why this is happening.
The number of bad candidates a recruiter has to filter is unbelievable.
4/ The people reddit/hackernews/twitter are competent programmers and they have absolutely no clue how bad a bad candidate can be. If I show you the programming samples I receive every day, you'll think I'm making shit up.
5/ To fill *one* position, a recruiter has to start with *100* resumes. And 85 of them will not know programming *at all*, no matter what the resume says and what job the candidate is currently doing.
(Note: numbers are for the Indian software industry.)
7/ Forget writing a simple program, most candidates can't recognize a working program.
Our questions regularly leak on the internet and then there will be "answers" on interview websites. Some answers are correct, most are wrong. And cheaters can't even pick the correct one
8/ And no, these are not hard, Google/Facebook/Microsoft-interview level questions. These are fizzbuzz level questions for juniors and only a little more difficult questions for the seniors. It took me years of being in this domain to really accept that the situation is so bad
9/ What's the solution?
Having a competent engineer screen 100 resumes for a single position would be a monumental waste of time. And at larger companies that have to hire hundreds of people, even non-tech recruiters can't afford to screen resumes manually.
10/This is why automated resume filters are popular. There has to be a way to bring the 100 candidates down to 30 phone screens and 15 interviews. (Even those are high numbers, but such is life. With @ReliScore we promise to bring the 100 down to 4-5 with just the online test)
11/ As @GergelyOrosz points out in this thread, if you spend some time trying to understand the predicament recruiters find themselves in, you wouldn't find this surprising
They need some signal to find the needle in the haystack and these keywords are it
12/ One hopes that the later stages of the interview process are better—when actual technical people get involved. But even that is not a given. Sometimes that is handled by junior engineers without any training in good interview techniques, and they make other mistakes.
13/ But, for most companies, the bottom line is that as long as the last interview (usually with a senior, sensible person) is good, they can afford to have inefficiencies in the earlier stages of the pipeline. Trying to create a 100% accurate filter makes no economic sense.
14/ This thread also reminds me of a more general principle that more people should apply before social media mockery:
15/ Of course the situation is neither ideal nor fair. Lots of good candidates get rejected. (That's the price companies pay to bring the pipeline down to manageable levels.)
If you're a good candidate, you should try to bypass the process
My code for @ReliScore is slowly falling for Greenspun's tenth trap, "Any sufficiently complicated program contains an ad hoc, informally-specified, bug-ridden, slow implementation of half of Common Lisp" as we implement automated eligibilty criteria for candidates
Customers want to make candidates appear for different tests based on various criteria (branch, percentage, score on first round test, etc). And of course, I don't want to hardcode that. So I've ended up with a mini expression language.
And now there's classic score creep.
Never imagined my eligibility specification language would end up with variable declarations and scopes
It has all arithmetic and Boolean operations and conditional branching. No loops yet
Looking forward to the day when a customer requirement forces me to implement recursion
If a student in high-school is keen on a career in AI/ML, how should they prepare for it? What subjects should they learn, and what activities should they participate in?
What basic subjects should you be strong in if you're interested in a long-term career in AI/ML? The replies to the parent tweet show that there are many different opinions.
Here's a thread on the slow and steady method, which focuses on building foundations for the long term.
3. You should have done well in maths and statistics in school. Nothing fancy, no multivariate calculus. Just basic probability and statistics and basic calculus.
You don't have to be a star student in math, and you don't have to be a topper. Just enjoying the subject is enough
1. Ransomware attacks are increasing. Even small companies in Pune/Bangalore have faced huge losses, in some cases coming to the brink of shutdown because of ransomware attacks
This danger is not really appreciated by people, so I wrote an article about this with @rohit11's help
2. There was a severe fuel shortage all along the US east coast last week. This was caused by a ransomware attack on the computer of the largest gas pipeline company in the US. The pipelines were shut for 5 days, and they ended paying $5 million in ransom
1. There are videos floating around on WhatsApp implying that Bhramari Pranayama (yogic breathing/humming) can protect you from Covid. I would like to poke holes in that theory and hope that a real expert (@bhalomanush?) will also chime in.
2. The claims start by talking about SaNOtize, a Canadian nasal spray that releases NO (nitric oxide) in your nose and claims to be a "prevention and early-treatment for Covid-19". Then they point out that bhramari increases NO naturally. Thus implying bhramari prevents Covid-19
An aunt sent this to me for validation before forwarding. Since I want to definitely positively encourage this behavior, I put in a bunch of effort at researching this.
The Utah government paid $20 million for an AI software that scans social media posts and identifies criminal activity in real-time.
Due to some problems they had to start an investigation: is the AI racially biased?
What do you think they found?
I love this story, so a 🧵 /1
Before you read the rest of this thread, try to guess what the investigation found in response to the question: Is the AI algorithm racially biased?
I bet the answer will surprise you.
But let me tell the whole story first. /2
Banjo, was a software company that made Live Time, a software with the capability to "detect active shooter incidents, child abduction cases, and traffic accidents from video footage or social media activity".
1. "There is too much variability in our hiring process," says the business leader. "Depending on which interviewer they get, people with very different skills get hired. Your tool will help us standardize."
Supposedly the customer is always right, but this is where I push back
2. I feel like wisecracking: if our software outperforms your employees, you need better employees
But I know that's not true. They are usually good employees who need to be given better training
3. So I dig into what is happening in the interviews.
Depending on availability, a candidate gets randomly assigned to one of 3 different interviewers. And each of those interviewers asks questions based on their experience, background, and what they're currently obsessed with.