Back to blog

My Friend Asked for Help With Her Resume. Thirty-Six Hours Later, Three Women Had AI-Powered Dashboards.

May 26, 2026

A friend asked for resume help. 36 hours later, three women had personalized AI-powered job search dashboards with resume tailoring, career coaching, and guided onboarding.

aidashboardresume
Listen to this page 0:00
0:00

It started with an email.

A friend from a previous job asked if I could help her tailor her resume for a specific position. 

She had the experience but wasn't sure how to present it for this particular role.

A woman with blonde wavy hair working at a laptop in a warm home kitchen, sunlit window, coffee mug nearby
...

The First Version Lived in My Inbox

We started with a basic email setup instead of a dedicated interface.

I created a specific folder in Gmail just for her, and I set up JJ, my personal assistant bot, to monitor it.

When she emailed a job description into that folder, JJ would tailor the document and email the result back.

If she wanted changes, she'd reply directly to the thread, and JJ would adjust the text and send the next version.

It technically worked, but it was incredibly awkward to manage.

As anyone who has worked with AI knows: training a system on a new skill takes time and plenty of back and forth iterations before it feels refined. 

Because we were relying entirely on an email inbox, that natural refinement process quickly hit a wall. 

The Friction Point

She was trying to work on multiple resumes at the same time, emailing JJ with requests for edits on them. 


The email threads got confusing fast and it became obvious that the process demanded a better solution. 

That clunky first version served its purpose, though, because it showed me how she was actually using it and what would make the whole thing flow better.
...

Moving from Email to a Dashboard

I decided to build an app.

The system would take her background, the job description, and a set of rules about how to present experience honestly, and then generate a custom resume and cover letter.

It would highlight which keywords matched the job listing and flag any obvious gaps.

But once I had the core tailoring feature working, I kept noticing other missing pieces.

She needed a few things in one place.

JJ, the little robot assistant, sitting at a desk editing a resume on a computer screen

She also needed a way to train JJ.

All of this makes JJ smarter over time.

Giving Claude that context meant every tailored version after that came back sharper.

I realized I had to put all of these tools in one place, so I built a five-tab dashboard:

Then I decided to set up a background process to scrape the internet for jobs and populate her leads tab automatically.

Not every scraped lead was a fit, so I built an ATS screener into Claude Code that scores the match against each job from one to ten.

Anything below a four gets filtered out before she ever sees it.

Screenshot of the Job Leads tab on the Girls Job Search HQ dashboard, showing six scored job postings with fit ratings from 5 to 9
The Job Leads tab on the live demo. Try it at pam-dashboard-seven.vercel.app.
Designing for clarity

She isn't a tech person, so every design decision was filtered through one question: would she know what to do here without me explaining it?


If a screen wasn't immediately obvious, I redesigned it until it made sense.

...

Two More Friends Needed the Same Thing

Two other women from the same workplace were also currently job searching.

Instead of building three separate instances, I restructured the codebase so one platform could serve all of them securely.

Each person received their own site with their own data, job leads, resume tailoring, and private chat history.

Adding a new person to the system takes about fifteen minutes.

The system learns each person individually, remembering what kind of roles they prefer and which experience to emphasize.

...

Built to Run Without Me in the Room

The onboarding experience had to be clean, because they needed to use the platform without me sitting next to them.

I built an 18-slide guided tour that every user walks through before they reach their main dashboard.

It shows exactly what each section does using screenshots from the live site.

I made sure the copy referenced things they would recognize from working together, keeping it warm and personal rather than reading like a generic software walkthrough.

That tour does the same job a live session with me would do.

It meets them where they are, shows them what matters, and makes them feel like the tool was built specifically for them.

I also included a directions tab that they can use as a refresher on how the system works once they're inside of it. 

The dashboard navigation bar reading The Girls Job Search HQ for Jane, with tabs for Start Here, Get a Tailored Resume, Job Leads, Training JJ, and Directions
The Directions tab is always right there in the nav, one click away.

This part comes from something I learned a long time ago in management.

If you let yourself become the person everyone has to come to for answers, you turn into the bottleneck.

You spend your whole day repeating the same explanations, and nothing else gets done.

So my goal is always the same: set things up so the system is the resource, not me.

The guided tour is that idea in practice.

The quick start guide intro screen, reading Before you dive in, lets walk through your dashboard, with the five dashboard tabs shown below
The walkthrough every user sees before they reach their dashboard. Eighteen steps, about four minutes.

It walks them through all five sections of the dashboard, one screen at a time.

Near the top of each slide, the matching tab lights up so they always know where they are.

You can step through the whole tour yourself if you want to see how it feels.

The What JJ knows and what to forget screen, showing what the assistant remembers about the user and a way to delete any of it
Everything you tell JJ shows up here, and anything you want gone, JJ forgets.

You can see everything JJ knows about you in one place, and remove anything that doesn't belong.

Tell JJ you sold Girl Scout cookies as a kid, and it might start hunting for jobs selling cookies door to door. 😄

One click and that memory is gone.

The Get a Tailored Resume screen, where the user pastes a job posting and adds special instructions for how JJ should tailor the resume
Paste the job, then tell JJ exactly how to handle it.

When you find a job you want, you paste in the posting and add any special instructions for that resume.

That's where you can tell JJ how to handle a gap, which experience to lean on, or how to frame a role that is a stretch.

The Job Leads screen showing a tailored job with View Tailored Resume and View Cover Letter buttons highlighted at the bottom
When JJ finished, the tailored resume and cover letter were waiting in her Job Leads tab.

You can open either one and edit the text right there on the screen to take it from "good" to "perfect".

A tailored resume open in the editor, fully editable, with Download docx and Print Save as PDF buttons at the top
The tailored resume, fully editable. Click to enlarge.
The matching cover letter open in the same editor, with the same Download docx and Print Save as PDF buttons at the top
The matching cover letter. Click to enlarge.

When it looks right, you download it as a Word document or save it as a PDF, straight from the top of the page.

If you ever forget how a piece works, the Directions tab walks you through it again.

What started as one quick favor had turned into something all three of them could use.

By the time I came up for air, the work was done.

...
Photo of Pam asleep at her desk on top of a keyboard, surrounded by code editors, sticky notes, and a coffee mug, with the caption '36 hours later...' overlaid on the image
The actual email I sent the three of them at the end of those 36 hours.

From One Email to a Platform

A project like this is when I'm most in my element.

I get completely focused, with a clear picture in my head of how the workflow should feel and what it takes to get it there.

I built the whole thing in about 36 hours.

And yes, I slept in there somewhere.

Once you take out sleep, meals, and time away from the screen, it was closer to 16 hours of actual work.

Jane Doe's sandbox is loaded with sample leads so you can click around the interface without signing up.

If you want to see how the flow works, the demo is open and running at pam-dashboard-seven.vercel.app.

Working on something like this for your team? Get in touch →