I tried with the query “principal software engineer”

Keywords were not helpful and included company names

For example:

One of the keywords was “fidelity”

Another keyword was “using”

A third keyword was “disability”

As far as requirements - some of the items on the list were actually requirements but others were just descriptions of specific companies and some were fragments of sentences that made no sense out of context

For example:

> This is a great time to join us and be part of this journey!Certifications:Company OverviewFidelity Investments is a privately held company with a mission to strengthen the financial well-being of our clients

> It also guarantees that your data security strategy adheres to compliance and data privacy regulations

Finally the list of requirements was very long - I’d expect the 10,000 foot view to be summarized a lot more - this seemed like it might be the same as skimming the 22 job descriptions it was generated from

So imo you’ve got some work to do before this is ready for prime time

First of all thank you for taking the time and sharing your experience.

> 10,000 foot view to be summarized a lot more

This is so spot on and point taken. The idea at the results page was to have filters to allow users reduce the amount of results shown. But, yes you are right the requirements list returns `one mighty scroll`. I will have a think of ways to tame it.

> fidelity I had mixed results for company recognition from a NER perspective, and this keyword I am afraid is the product of that. I have some ways to reduce reliance on NER here, will see if that fixes it.

> disability these type of words are slightly more nuanced as in some cases these will be relevant keywords... this I will have to iterate more.

> sentence samples I expect the model to get better over time to weed these out. I am somewhat weirdly confident about it as I have seen 1st hand how it missed obvious things in earlier versions, but is doing much better now.

all in all thank you once again for taking the time, have a great day.

I didn't understand what it was meant to do. I think you should have a clear explanation of how to use it on the front page. For example, what job description should I be searching for? And when I do, what will it tell me?

I tried searching for "Data scientist". I got what looked like snippets from existing job adverts, also a salary range. I'm not sure what I would do with this information to update my CV, though.

Thank you for letting me know, this outside feedback is actually helpful.

The main idea of BDDB is to get an overview of job requirements all in one place. Snippets you found, are there to help you: a. highlight relevents skills in your resume and/or, b. evaluate what training you should plan for career advancement.

really fair point about making the homepage more informative, my view was to keep it aesthetically clean.

the search function is mainly good for terms as they would occur in a job title.

Good luck!

Wow 450 hours! I bet you learned a ton about all the different layers that went into this project, good for you on the undertaking.

You’re probably on the right track here, the execution may need a few more iterations. Ultimately if you average out a bunch of resume descriptions, the average may not end up being what anyone wants. It’s like the idea of an “average height or size” for humans, where the clothes are mostly made for one size, but the reality is that everyone is all over the place.

Bad analogy aside, I would find usefulness in having a program that has an input of a resume and a job description and then outputs a new resume that has a base from the input but that is tailored to the job description, using AI or whatever.

Strange feedback. 2 pages is fine.

In my opinion, people with more than a decade of experience should feel fine having a 2 page CV.

But anyone with <= 10 years of experience should be able to summarize their accomplishments well enough for it to fit on a single page.

Thanks I agree, I might have slightly over-reacted.

I don’t understand what this site is supposed to do. A resume is generally one page; an academic CV is exhaustive and cannot be collapsed into one page.

Cool work! On a similar vein, this would nicely complement LinkedIn's Career Explorer [0] to make it easier to write a CV. I hope more freely available tools on this gets showcased in the future.

[0]: https://linkedin.github.io/career-explorer