Show HN: HN Resume to Jobs – AI Powered Job Matching Tailored to Your Resume | Hacker News Reader

Show HN: HN Resume to Jobs – AI Powered Job Matching Tailored to Your Resume

hnresumetojobs.com

214 pointsSCUSKU75 comments

Hey HN! I'm excited to show off this side project I've been working on. This project matches your resume with the best matching jobs from the monthly HN Who's Hiring post. It works by creating a vector embedding of your resume using OpenAI's embedding API, and then ranking the jobs using a vector similarity score. (You can toggle between max inner product, cosine, and euclidean in the "Advanced Options")

I was laid off in August and it took a whole 6 months for me to find my new job. Fortunately, I found my new role on January's HN Who's Hiring post. So I hope this will prove useful to any job seekers out there. I know it's a tough time right now, but you will get through it!

Thanks HN! I would greatly appreciate any and all feedback!

Open articleView on HN

That sounds like a good idea, but I do wonder if it's a little too... safe?

Every job I've ever taken, there were bullet points in the requirements that I didn't have. And every job has been in a different industry. (Granted, this is only 5 job hops in over 15 years) I'm not really sure I'd want to be perfectly qualified for a new job. Part of the fun is learning new stuff.

Of course, I do understand that any job is better than no job, and it's much easier to land one where you do hit all of the bullet points.

This is a really good point. I've also moved upwards into jobs I don't technically meet the requirements in most of my roles. I wonder if an AI tool could generate a new resume based on your existing resume and the job spec for a role to find the best intersection of the two in order to increase your chances of getting an interview...

This is exactly what i have been working on for few weeks but not been able to get a good final results. I want to make sure that AI is not overselling my skills or completely deviating from my current experiences/skill set when trying to tailor resume for a job post.

This is an interesting problem, I might start by having ChatGPT try to infer a complete list of tools and skills each position involved along with a confidence score (even those you didn't originally tell it about)

From there you could allow it to remix the skills while maintaining an average level of confidence greater than some number: so it'd be allowed to include things slightly out of your wheelhouse only if most of the job skills matched up perfectly.

You wouldn't need the math to check out perfectly there, the score would just be a heuristic it could use internally

Cool, i should try it..so far i have been working with rephrasing the work experiences description to fit a job description using GPT3.5 api. Still have not engineered the prompt to get reasonable output.For tools and skills , i have a prefilled dictionary which more or less represent my expertises. task of GPT is select top n skills/tools and put them in final result.

Another challenge is to get the AI generated output to fit in a page and preserve original formatting

yeah using chain of thoughts to tackle it. This is relatively simple

The idea of using AI to tailor a resume to a specific job spec is promising. Although there isn't direct research on generating resumes using language models, I found related studies in MirrorThink.ai that provide insights into this application. For example, [1] demonstrates the potential of using prompt-based learning for extracting information from resumes with only a few labeled examples. Another relevant study, [2], explores the tradeoff between removing gendered information and maintaining useful job-related information in the resume screening context.

These papers suggest that AI models could be employed to analyze your existing resume and the job specification, and then generate a tailored resume that highlights the most relevant skills and experiences. The main challenge lies in striking a balance between tailoring the resume and maintaining an accurate representation of your abilities.

[1] A Few-shot Approach to Resume Information Extraction via Prompts - 2022: https://arxiv.org/search/?query=A+Few+shot+Approach+to+Resum...

[2] Degendering Resumes for Fair Algorithmic Resume Screening - 2022: https://arxiv.org/search/?query=Degendering+Resumes+for+Fair...

Is this not a matter of training? once you know career moves on resumes and how experience can provide the next jump in role it'll then isolate those roles for you. same for cross industry successful moves. I'm sure this is how LinkedIn does this although it's never perfect.

recent advances would allow a platform to conversationally review your CV and probe you're desires to help with recommendations. super interesting space.

AI is statistics based. It's not going to give you only 100% matches.

> Every job I've ever taken, there were bullet points in the requirements that I didn't have.

This. And you're like most people.

Most job descriptions and requirements lists are describing the "dream candidate". Very few employers actually expect anyone to have every requirement in the list.

What I'd love to see is a view on the "who's hiring" where I can start by saything something like:

* I live in Germany, and want to work remotely.

Then it filters out all the onsite jobs as well as the remote US or remote UK only jobs. It should be smart enough to know that Germany is in Europe, so Europe or anywhere remote jobs should remain.

Then I would like if it could extract some features and show them for the remaining jobs, and ideally I could filter by those features.

Features I'm interested in:

* type of employer (government, corporation, startup, non-profit)

* ethical orientation (wouldn't want to work for an ad company that sell user data, would very much like to work for an EV charging startup that tries to contribute to solving climate change). If that doesn't work, industry might serve as a proxy

* technologies used

* role type (developer, architect, product owner etc.)

* salary range (if available)

Yeah, when it said 'show distance' I thought it would offer me jobs nearby/remote, but instead it's talking about an algorithm distance which probably nobody outside that specific domain really cares about.

As for EV related jobs, what I have been doing is checking as many related conferences as I could find and EV related channels in YouTube and visit each one of the homepages for the companies. For conferences it can take up to 4h, but it's very quick (but tedious) to remove chaff that you're not interest into.

I'm haven't been lucky, not many are in France, very few are remote, but I have seen quite many in Germany, so maybe you're in better luck.

Also, https://climatebase.org/ and maybe another reread of https://news.ycombinator.com/item?id=26301765

Does selecting the "Germany (remote)" filter here address your use case? https://www.trueup.io/jobs

That looks pretty good, thank you!

I don’t feel safe uploading a cv to an untrusted site. It doesn’t explain what happens with the cv on first glance after it’s uploaded.

Good luck

This, and it kind of bothers me that that is my first thought on the internet these days. I get enough unsolicited mail/phone calls from recruiters as it is.

I appreciate the well wishes, your suspicion completely understandable (and frankly warranted). Others in this thread have suggested a Privacy Policy, looks like that's the next thing on my todo list. But as it stands, the resume text is stored, while the document itself is not. This is so I can cache resume embeddings to try and keep my OpenAI API costs to a minimum.

Stored for how long?

I didn't know this about ChatGPT - brokers of user data, trading data collected for compute. As interesting as this is, there's a lil bit of an interesting public policy and computer technology for fair use of one's personal data. No expert, but might be a slippery slope long term, so highly suggest you get a checkbox in there with some sort of absolvement in liability between your app and the user's resume.

Other than that, it's like all good things - 80% of it is the button click, while the last 20% is proofing the final result and correcting the model so the machine can learn.

God Speed man. Love new tools.

Same here. So I just made a bullet point list of the technical things I have done in the past. I got a list of jobs I wouldn't necessarily have searched for on a job board so that's nice.

? Personally I'm comfortable uploading my resume wherever - it's my resume, it exists to advertise my skills. I'd give it freely to anyone who asks, there's nothing on there I'd consider private almost by definition.

Surely your CV is inherently public information, isn't it non-negative for you if it's shared or 'leaked'?

Surely your CV is inherently public information

Says who?

I find it actually quite strange to think of not only your own, but everyone else's professional (and implicitly: social, geographic) history as "inherently public" information.

Says you the author of the CV without attached terms limiting distribution? You who wants people to read it and want to hire you?

Pure curiosity.. what evil could come from your work history being public or falling in the "wrong hands" whatever those may be?

Unsolicited calls and emails.

You already sort of have that with a public profile. And you don't need to include your phone number. You basically don't get my phone number without a specific reason.

I've only gotten jobs through people I've known since the 1980s but at least a capsule version on my resume is embedded in bios on countless event sites. It's hard (and undesirable) for a lot of people to keep that information hidden.

Somewhat off-topic but congratulations, you have re-implemented (in the small) the search engine that powers HN. I say this as an employee of the company behind that engine, and I say this as someone who is intimately familiar with the codebase and with how the technology works.

With a little more effort towards scalability, and with some input from a devops co-founder, you'd be on your way to launching a state-of-the-art, market-competitive search engine.

My employer won't be too happy seeing me post this, but I don't care, this needs to be said.

Whats special about this particular cosine similarity search? There are loads of these! Like every other day on HN since chatGPT was released. Several PDF chatbots. There are various overfunded vector search companies. Some are open source. And then there is elastic search etc.

We are not at all in disagreement. In fact, you are making the same point I'm making, which is that there is nothing special about any of the commercially available search engines today. With access to OpenAI's embeddings API and to a vector database, a mid-level engineer can build a highly scalable search engine in a few weeks. As a startup, it makes sense today to build your own search engine rather than buy off the shelf.

The only thing companies like ElasticSearch and Algolia still have is their pre-existing customer bases, a few thin layers of marketing, and some network effects. Search engine companies are effectively marketing companies nowadays.

That explains why the corrupt management at my company thinks it's a better strategy to throw hissy fits on social media in the general direction of OpenAI, than to work on actually building useful tech.

I see what you mean. Interesting topic! Some startups will just want to move fast and use their funding to have someone else solve it. I work somewhere where we use Azure search and honestly I don’t think anyone has event talked about it for years. Just sits there doing it’s thing! And not having to maintain that and getting feature enhancements automatically (double edge sword yes) is good. You made that same point though, we are an existing search customer. But I am sure if we were building again we wouldn’t roll out own. The problem Algolia will have is the immense competition. Why leave your fuzzy favourite cloud for search?

When I was headhunted for my current job, I asked him how he found me. The agent then told me that they had an AI system for finding and comparing candidates. He then reassured me that I was well qualified for the job he was pitching to me. So, I took the job, and I've never been happier!

Now, I wouldn't call my formal qualifications exceptional for the tasks I'm doing right now. You both need intimate tech knowledge and organizational skills, on top of some very specific knowledge about systems, algorithms, and AI. The width of tech you need to know for this job is kind of extreme, and IMHO that's not easy to read from my formal resume which mostly shows a background from teaching.

I do however have a personal tech blog and some projects on GitHub. I wonder how well their AI tracks such extraneous links, because if I was a tech company I'd first and foremost want to know specifics about what I know, how I apply that knowledge, and how easily I am able to learn and pick up new technologies. I think an AI is the perfect tool to automate such an analysis, so I'm wondering if that's what they did.

I guess opening Pandoras box would be to also employ forensic text fingerprinting to find other projects from the same author under various pseudonyms.

Great work and congrats on the new job!

If you have a backend, perhaps add a feedback loop from the user which ones are relevant and which ones are not.

Using OpenAI, perhaps you can analyze the attributes and allow user to filter by Location, Title, size of the team, tags etc.

Thank you! It's been great so far, and I actually learned about some of the technologies in this project from my new job (RTKQuery to be specific).

I actually do have a "Apply Filters" button, have you given that a try? I do thing the team size tag is a great idea.

After clicking search, the scroll takes me to the first result hiding the number of jobs found and apply filters and I missed it. :D (A nice to have would be perhaps, if you apply a filter, you highlight the word of the filter applied.)

I don't think the user of the post is the most important of the post, so I would use another color to not make it pop up.

Curious how this matches as I put my wife (attorney) cv into this and it kept giving me product designer or front end engineer roles (none of which have anything to do with her resume.

> This project matches your resume with the best matching jobs from the monthly HN Who's Hiring post.

I don’t think I see many attorney listings in those threads.

Great that I can enter free text to enter relevant information without needing to upload the resume.

I entered comma-separated list of keywords I'd like the job post to have and got quite nice results.

Very cool, but it actually works better if I just use the CV field as a search field and type what I'm interested in working on.

I did not anticipate that behavior, but I'm glad it's working well. I will add that to the placeholder text in the text input so that others are also given that suggestion. Thanks for the feedback!

You need a privacy policy!

I'll second this.

I would like to give this a try (been job hunting for the past 6 months), but I'm also a bit paranoid in sharing a lot of data, especially with a service that has no privacy policy.

Very fair point! Not to put the onus on you but do you happen to have any pointers on where I can get started?

As it stands, I am using google analytics, and also storing the resume text (though not the document), and the embedding of the text. I hope at the very least that provides some transparency.

Obligatory disclaimer IANAL (I am not a lawyer)

Using Google Analytics might have some GDPR implications in Europe. A more GDPR-friendly alternative might be Plausible Analytics.

Also, given the amount of information contained in a CV, even without a name, address, contact info etc, it could make the process of identifying the actual person fairly easy, thus it might have some GDPR implications.

You could have a look at some Privacy Policies from recruiting agencies [0][1], though those are (in my opinion) fairly "abusive".

my 2 cents: given this is a side project, I would not store any data at all

[0]: https://www.greenhouse.com/de/privacy-policy

[1]: https://join.com/privacy-policy

https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-re...

The ICO has some good guidance. They’re your regulator in the U.K.

Get privacy policy from iubenda. Easy and cheap.

Use Wide Angle Analytics instead of Plausible or Google Analytics. Much better choice if you ever leak personal data to your analytics platform and very few solutions support PD. Wide Angle does.

Given the personal data you are processing, some might even become sensitive (disability status), you need DPO ( Data Protection Officer). There are DPO as a service out there.

This is really cool! I've been toying with trying to make this as well. What are you using for parsing PDF's? I've found most libs are pretty bad at extracting content well from PDFs. If anyone else has any recommendations would be much appreciated!

I am using a tool called pdftotext (https://www.xpdfreader.com/pdftotext-man.html) here. I have used it in a previous project with great success when other tools would bug out or return a bunch of weird junk text. Give it a try!

Cool idea, congrats on launching! Though I ditched resume a long time ago to only use linkedin. Having the option to automatically fetch info from Li would make it easier for me to try.

LinkedIn has a resume builder option now. Or you can just download your profile as a PDF.

Was thinking the same, if this could extract my profile from LinkedIn.

This does not make sense to me. First, you might not want your next job to be just like the ones on your resume. Second, you might have certain constraints, and value certain things more than others in a position; i.e., have your own ranking function. Third, this site is of no use without job listings, and unless you have your own data, you can't really do anything commercial.

But congratulations on your new job!

Not bad. Got some possible roles in the first couple of pages (enough that I actually clocked on the links). The ones that missed were reasonable too, at least as good as searching for keywords on my local jobs site.

I'm Operations so not a huge number of roles on this site though. So it might have effectively been giving me everything with "SRE" or "devops" in it.

A friend of mine and I explored a similar offering. We've mostly put it on a shelf as it was very duct taped together and we're both short on time. We don't have any monetization plan but would love to hear any feedback: https://divim.shovelshop.io

Seems pretty accurate. Based on my (almost true) resume[1] the top result was:

  Man Group Alpha Tech | Quant Python Engineer | London, UK

[1] Resume:

  male
  sharpshooter
  can run 1 mile in under 6 minutes
  bench press 300 lbs
  IQ of 137
  7/10 attractiveness
  10% bodyfat

  :D

This us a great tool!

One thing that would make it a lot more useful IMO would be if the posts were somehow collapsed at first, showing the headline information about the role, so I could open the ones that look interesting or sound like a good fit instead of needing to scroll past their full text.

Resume: I am God. https://hnresumetojobs.com/?resumeUUID=1495b8a8f85b463baa3fe...

I love that the first reply is for Minion.AI. Are you a god? Do you want minions? We can get you minions.

This looks cool. I use LinkedIn as my most up to date resume. I could export to PDF and then upload it to you, but it would be great if I could just put in my LinkedIn URL.

Generally speaking, this is an interesting application.

Users will tell if it's useful, but regardless, I'm curious to apps that use LLMs in such ways... Returning lists.

Awesome work! I would use this and I personally know a couple of people who would too!

Are the jobs up to date? How are you sourcing jobs? Are you searching indeed.com?

Hope these kinds of toolings will kill al HR Agencies.

I am building something very similar actually. Would it be possible to connect with you sometime?

Sure! Email is in my profile :)

Very I interesting project. Thanks. I’m going to start using it for my job search. Great job!

Thanks for checking it out! Hope you find something great, all the best on the job search, it's a tough process but something will work out for you :)

How does this differ from LinkedIn which does the same thing and has huge network effect?

No feed of humble brags is a feature.

OMG 6 months, I like to hear about that too. The job search is a struggle.

Can I leave a nota at the site hnresume?

Maybe just my CV but none truly fit

Honestly I've found that with my resume as well. Like others are saying, there truly is no perfect job. That said, perhaps you could try searching previous months and see if anything good comes up that way?

I imagine it could generate a lovely list of things one might consider learning.

My skill set is (css,) vanilla front end js and enough php to dumb db queries down the tube into the front end. There is no market for my write once works forever. I'm not interested in learning anything else but it would be wild to see an ordered list of suggestions accounting for marketability and difficulty. No way a puny human like myself can make sense of an n dimensional universe.

can I leave a note at HNResume

Reply on news.ycombinator.com