https://ico.org.uk/for-organisations/direct-marketing-and-pr...
- The majority of the page is wasted on irrelevant quotes and then explanations about irrelevant quotes. AI is all around us theres no need to sell AI.
- "Design and Development Choices" This section should be moved to the "about us", anything that isn't about the product shouldn't be on the landing page.
- There aren't any actual examples of what the tool does, what it looks like, why i need it.
People are using our AI-based tool to find and apply to jobs.
[insert spiderman meme]
You are very likely providing a thin layer on an LLM which does summarization based on criteria. It would be reasonable if that's either a few-shot approach ($0, one time) or a fine-tuned model ($6 - $9 per million tokens, one time), plus the running costs ($0.0001 - $0.01 per run) for llama2 3-70b or gpt3.5-turbo-instruct.
Is there any additional USP, like have you used a data set sourced from hires in specific market sectors to know what parts of a resume really enhance or hurt the chances? A similar study on ATS/real filters would be also be a USP. Or another USP could be that it's super easy to use - drop a resume in, Apple Pay $1, it's done. I'm not seeing a lot of value if I can discuss my resume with ChatGPT and other assistants for free to get a second perspective.
Best of luck, of course.
To get structured data, it's not just about parsing the file (PDF, Docs) with an open source to get the text. You need to extract skills, qualifications, work history ... etc.
There are tools already doing this part very well (before LLMs), so you need to make multiple LLM calls to make the right match or provide useful information.
Your comment suggests that you think it: Hey ChatGPT, this is my CV, and this is the job description; review it and give me suggestions.
Building systems around LLMs is not that easy.
I build AI products in a known company and experiment on my own time. It is really easy to build a product like this without aligning it to solid research or recruiter ATSs. I know the value of this alignment because part of my job is interviewing candidates for my team. Easy isn’t bad, of course.
The tool above is undercooked. It’s a very cool week-long project or a nice back burner one. I have nothing against that, good on GP for building it.
My point was that the presentation and pricing was a bit aggrandizing. Given what the project is, which is indeed similar to running a query with ChatGPT on a resume as it mainly does structured prompting, it has a better shot at being a $1 product. For more, it needs some of the things you and I mentioned.
There may be further value in providing ATS capability to recruiters — a tool that maybe sorts hundreds of resumes by various criteria a recruiter inputs. Such a tool could also highlight blind spots — “many candidates mention X, Y, and Z”, “This candidate stands out because A and B”. Such work would need to be done very responsibly, however.
