Hacker News Reader

Resume format is definitely a big reason here. I used to have a resume that looked like this – two-column, made in Adobe Illustrator, fancy icons/fonts. I applied to 150+ places with maybe 5 callbacks. I switched over to a super basic .docx template and got 3x the callbacks.

Resume aside, the job search process itself is also insanely outdated. Companies still force you re-enter your resume on their hiring portal and then have recruiters manually review resumes. We're building something like a "common app for tech jobs" here at Simplify (https://simplify.jobs/) to help candidates find roles that actually fit them and recruiters effectively process inbound. Praying for the day companies stop using Workday as an ATS....

Disclaimer: I am the founder (we’re also YC backed!)

@mikeyan320. Curious, how is simplify.jobs handling the labeling/disambiguation of skills and platforms? I've been working in this domain for a few years and have identified this problem as a significant obstacle. I open-sourced struct-ure/kg, a knowledge-graph of IT skills and platforms, as a possible solution: https://github.com/struct-ure/kg. I'd welcome your thoughts on the subject.

oof, no offense but I wish you guys didn't push your browser extension so much. that comes off as incredibly suspicious behavior in my book.

Yikes. PDF without any of the text embedded. Can't select, can't search, can't filter. He's basically sending everyone a JPEG!

I would expect this to totally not work. Maybe someone is going to have OCR somewhere in the chain, but... why would you bet on it??

When I was applying, I had a situation where sometimes the ATS system would show me as Geoff and other times as Geo, with no clue why.

Obviously, that's not a fatal issue, but it was indicative of other problems with my résumé, and I saw those in the auto-generated job histories.

Pixelmator Pro just released a feature where it can open PDFs to show you the internal structure. I opened my résumé and sure enough "Geoff" is in two text boxes, despite being a single chunk of text in the source document. :-/

How did you create the PDF? Sounds a bit like the ff is a ligature (i.e. not simply two f’s, but like one special character consisting of two f’s)

https://en.m.wikipedia.org/wiki/Ligature_(writing)

"ff" can be a ligature,ﬀ. It is possible their system doesn't like the Unicode character. Even if isn't ligature character in PDF, the renderer could make it. Decomposition errors are common, as are not handling Unicode.

It's not a ligature, but you're definitely onto something. I checked again, and all the characters of my name are individual paths, except for the f's. I put a space between the f's, and now they're two separate characters as well. I double an n, no grouping. I put in a double f elsewhere, the grouping happens for those.

It's amazing to me that the PDF ecosystem is this broken after thirty years. I'm boggled that there isn't a commonly-used way to add structured data to a PDF to enable it as a machine-communication standard as well as a human one.

Edit to add: I found a "disable ligatures" option in Pages. Setting that didn't fix the "f" grouping. :-/

There’s a way, it’s just not commonly used. (See: PDF/A).

To be fair most ATSs are really bad at parsing information in very unexpected ways.

I used to have a pretty basic resume with two columns, basically plain text describing job experience and education in reverse chronological order and a narrow side column showing experience in languages and frameworks in a progress bar style to indicate proficiency/experience.

Some of the application software of the “fill everything in regardless of resume” variety offered to upload resume and then auto populate, which is essentially a preview of the ATS parsing capabilities.

It was perfectly capable of parsing the experience in languages etc, but would consistently fuckup the plain work history.

Even recreating the resume in a simple no frills Word document would help.

In these cases I was lucky I could manually input everything, but it did leave me very unimpressed.

Any good resume submission system should be capable of OCR.

But what is this guy? Sounds like he's a designer, a communications specialist, team leader, and project manager. Just all over the place. I'm not sure what I'd hire them for.

Edit: why disagree?

Are they actually good though? (the ATS systems) I've only recently learned how important they are in general and I'm not familiar with their capabilities.

I would think that even with OCR, this resume might be confusing because the years and contact information and stuff is kinda spread around.

If someone has has good insight how well/not well they work, I would appreciate if you commented about it.

There will always be some edge cases where it can't identify or interpret something, but they're pretty good. It doesn't matter where something is located in the document, it can pull stuff out like an address or phone number based on format and stuff.

And yet they don't. When I was applying some systems got my name wrong. They would chop off two characters. I assure you, that's not how it looked in the PDF.

If they're expecting text, then why even submit in PDF? Wouldn't a Word document be better? At the very least, if a system has limitations or and expected rigid input, then it should provide that information to the submitter. I'm sure there are terrible systems our there, but I'd be surprised if the average ones aren't decent at OCR.

Be careful, the wrong settings, and all your edit history may be available.

(This could be bad, depending...)

As well, as some others mentioned, some sleezy recruiters will change your resume without permission. Add things/skills. Remove things.

These were systems that were happy to take word, text, pdf, and a few others. Often they expressed a preference for pdf.

A word document is also a binary format (unless it is an xml monstrosity) not too different from pdf, but more proprietary. Plus one vould embed an image on word.

With a markdown format I would agree, though.

> Edit: why disagree?

Because I don't think a resume submission system should be doing OCR. There aren't that many text formats out there: between Word, PDF, and some form of plain text (just actual plain text or something like markdown or readme), that probably covers 99% of resume formats. Why on Earth would I want to run resumes through OCR when all those other formats can easily have their text extracted?

Even if I'm sending it in PDF, I at least want it to be such that in case the document I've uploaded does wind up on some HR mook's plate, they can actually select, copy, and paste text from it.

That's fine if the system discloses to the users that it must be in a text format.

When you submit a text document, the text should actually be text. This is the most basic step.

I would not want to hire a designer who sends text as an image.

PDFs are not necessarily text documents. That's why it's important for the system to specify if it's expecting text, or use OCR.

No, but text is text. It is perhaps easier for the designer of the document to just put some text from photoshop rather than tweak the document, but it’s absolutely user hostile.

100% agree. this thing is WAY too stylish. just put the information so that a computer can read it in. the annoyance factor is probably enough to reject him in some cases.

Oh come on. OCR is a thing. I mean this is HN, are you normalizing acceptance of crappy software?

I would not blame the software. I'm pretty sure that images are not generally used as formats for resumes.

I mean, at what point do you think it gets ridiculous? When it's a video instead of an image? When it's in Sanskrit? (Computer translation is a thing, after all.) When it's a video in Sanskrit?

Or maybe is it more reasonable to say, look, this is some text, submit it in a standard text format? And an image is not that, even though OCR is a thing.

I agree that what you said is more reasonable but most jobs are not for technical people. Even somewhat technical graphics designers might go over the top and craft a visually appealing resume and use a picture format.

Why blame anyone for this? Just use OCR where possible. If the application instructions specify a text resume sure. Having a picture resume in a .docx makes little sense but a PDF is both an image and text format. Not to mention a lot of people design their resume or have someone else make it and print it for them and scan it when applying for jobs as an image pdf.

Shit happens. Tech is supposed to make life easier not harder, despite of humans' failures not because humans arr cooperative enough right?

Why bother with OCR as you still find enough people without it. And if you don't get enough applicants that you need automation manual review will happen. So still you don't need OCR.

2000 applications is quite rare. A posting is "hot" on indeed above the 50+ applicant mark from what I have seen.

If you have under 150 resumes there is no reason why a human can't skip+prioritize a few times and read in detail the top 50 or so.

A liberal 4 minutes per resume is 15/hr,120/8h. You can't spend or have someone spend one day reading resumés? In my earlier example of 50 resumes, that's less than half a day, a morning reading resumes to decide who you will follow up over a call in the afternoon.

Corporate laziness is such an epidemic. Honestly, why shouldn't ML replace this part of your job then? Let ML do the interview and hiring too. This middle ground where lazy keyword oriented algorithms and heavily biased humans (goof luck if your last name is washington!) is the worst of both worlds for everyone but the managers and HR people cruising their way lazily to retirement.

"You can't spend or have someone spend one day reading resumés?"

Managers are even lazier than devs.

One thing that is worth keeping in mind is that if a team is hiring they're probably understaffed, which likely means that they don't have much spare time. Hiring should be a high priority in that case but it's often a struggle to even get a small amount of time to focus on it.

That's what managers are for, this is what they do, not the team. They're supposed to delegate actual work to their team and spend time on this sort of a thing so that their team has the right people and resources.

I would expect no less from a JP Morgan employee ;-)

I got the advice to imagify all the text in resumes, because some recruiters would 'upgrade' the text before sending it to companies. Next thing you know, you have 5 new skills for things you've never heard about that somehow happen to perfectly match whatever $random_corporation$ needs.