As a consultant, I have to submit my resume a lot. You always want you resume tailored to the project you’re trying to land. In the old days, that meant taking the most recent version, tweaking it, and hope that any gaps from the previous version aren’t fatal.
But these aren’t the old days and we have ChatGPT. So about two years ago, I took a weekend and wrote a detailed narrative in prose to describe my 30-year career. Since then, I have kept the narrative up to date, by adding paragraphs describing my work and accomplishments. I then use the narrative with ChatGPT to do things like assessing my qualifications against contract or labor category requirements, extracting resumes that highlight specific skill profiles or align with formatting requirements, identify specific gaps in my skills compared to requirements, or any of a number of such tasks. Let’s take a look.
First, let me introduce you to Alexa Stone. Alexa is a fictional Senior Geospatial Data Scientist who was dreamed up by ChatGPT at my request. I asked it to use my narrative as a guide for creating a fictional one for Alexa. It can be found here.
I started a ChatGPT session and set some context by telling it to be an HR Consultant. I then uploaded the markdown version of the Alexa’s narrative.
It’s a good idea to force it read the document by asking it to describe the contents. If you feel like it still hasn’t processed it thoroughly, you ask it to count the words in the document.
That looked good so I moved on. I then asked it to produce resumes for Alexa – one tailored to highlight leadership skills and the other to highlight senior technical skills.
The results can be found here and here, but the follow screenshots highlight the different takes on her most recent role. Here is the leadership version:
And here is the senior technical version
The differences are subtle yet also distinct. ChatGPT did a decent job of tailoring her work and accomplishments for each case. It’s easy to see it’s the same work, but that each take has a viewpoint appropriate for the role being sought. There’s a big caveat to keep in mind, but I’ll get to that at the end.
That’s fine for generic resumes tailored to skill profiles, but what about one that’s being developed for a specific opening. In this case, I went to LinkedIn and found an actual opening – a Solution Architect opening at Voyager Search. The LinkedIn opening is here, but I saved a PDF for posterity here.
I uploaded the PDF to ChatGPT and asked it to assess Alexa’s suitability for the role. Here is what it came back with.
It did a decent job of highlighting gaps and show how her strengths could address those gaps. For example, she doesn’t have direct experience with Solr or Elasticsearch, but her cloud and data pipeline experience probably give enough of a background that she could come up to speed. This is relatively decent guidance, especially from a machine.
So now let’s ask it to make a resume that is aligned to the Voyager opening. The full resume can be found here, but here’s the excerpt of her most recent role.
As you can see, it made some attempt at alignment. It tried to address the transferable experience question with “such as Solr and Elasticsearch.” I’m not in love with that approach, which is getting us closer to the big caveat I mentioned earlier. I think this does highlight in the resume where we can discuss the transferable experience and begin to make clear that Alexa is really qualified, even without direct Solr/Elasticsearch experience, but edits are needed.
So that big caveat: ChatGPT is a tool. It is not a replacement for your brain. You can iterate in ChatGPT and get pretty close, but you will need to eventually paste it into a Google Doc and do the last-mile editing. The example above highlights this. It still strongly implies direct experience with Solr. I would manually edit it to soften the language and make it clear that Alexa has experience with related tools that make her qualified. Sure, I could probably iterate with ChatGPT to get what I want, but diminishing returns kick in and it’s more efficient to do the final edits myself.
In summary: Never take the output from ChatGPT and simply ship it without a critical review.
Resumes are good, but there’s more we can do. With Voyager opening loaded, I asked ChatGPT to produce sample interview questions to help Alexa practice. I asked for 10 and this screen shot shows the first four. Some of them were a bit broad and I could refine them, but they were still useful for practice questions.
Wrapping up – I’ve been pretty happy with the results I’ve gotten from ChatGPT when I’ve used it for this task. (Yes, OpenAI knows a lot about my career now.) Large language models are pretty good at processing language, so a prosaic career narrative is a great data source for this type of work. It also has the added benefit of being pretty easy to maintain. I simply open mine up every couple of months and add to it. I am creating my own corpus that I can pare down and customize with an LLM.
I can imagine LLMs having a lot of utility in the government services space where formatting resumes, aligning them to labor categories and key personnel requirements, and then doing it all over again for the next procurement can be a huge time sink. My next step is to pull this all out of ChatGPT and into a local LLM for just such work.
The takeaways: First, ditch your resume(s) for a career narrative. Take some time and write it out. You’ll thank yourself. Second, don’t just take what the LLM gives you – review it and edit it before you ship it.
Resume tailoring and maintenance is as much art as science, so I hope this approach can be useful. Thanks for reading.
