Laid off after 11 years, they built an AI assistant that checks its own work
An AI Career Assistant that screens jobs, tailors applications and runs a script to verify its own accuracy.
After 11 years at one company, Monika was laid off just as the job market shifted. Instead of drowning in scattered, manual searching, she used AI Mastery to build a personal Career Assistant in Claude that scores postings, writes in her own voice and audits itself. It is a clear example of turning a hard moment into a working system.
M
Monika Jaszi
A professional navigating a career transition after 11 years at one company.
Before
Monika had spent 11 years at one company when she was laid off, right as the job market changed. She needed a way through a career transition without burning out on scattered, manual job searching. Every posting, resume and cover letter meant starting from scratch.
“After 11 years at one company, I was laid off right as the job market shifted, and needed a way to navigate a career transition efficiently without burning out on scattered, manual job searching.”
In their own words
The turning point
AI Mastery gave her the confidence and a framework to build properly. She paired a living Project doc, holding her background, voice and pipeline, with a Claude Skill that turns it into repeatable workflows. When the same errors kept resurfacing across files, she wrote a script that checks the whole system for accuracy on its own.
“Being part of AI Mastery gave me the confidence and the framework to build this properly, from the structure of the skill itself to the visual case study.”
In their own words
After
Monika now runs a Career Assistant that screens postings, tailors resumes and cover letters, preps her for interviews and tracks a growing pipeline. She stays hands-on, reviewing every output and pushing back on suggested fixes when they do not hold up. She has screened 23 real postings, applied to only 6, and two advanced to recruiter screens.
What it lets people do
The assistant lets a job seeker rule out poor-fit roles before spending time on them and focus energy on the ones that matter. Its consistency script catches drift in facts, dates and statuses automatically, so the system stays trustworthy. Monika is exploring a template others could adapt for their own search.
What they learned
Monika learned that when information lives in two places, something has to enforce keeping them in sync, not just memory. She also found that a checker can be wrong too, and must be tested against real files and read critically.
“I learned that a checker itself can be wrong, and needs to be tested against real files and read critically rather than trusted on first output.”
In their own words
See the work
The sync loop: project doc, skill files and consistency script
The Career Assistant Skill was built during AI Mastery.
Read the full submission
Problem Statement: After 11 years at one company, I was laid off right as the job market shifted, and needed a way to navigate a career transition efficiently without burning out on scattered, manual job searching.
Achievements: Built a personal AI Career Assistant that screens job postings, tailors resumes and cover letters in my own voice, preps me for interviews, and tracks a real, growing pipeline. It's backed by a checking tool that verifies the system's own accuracy. Since building this, I've screened 23 real postings and applied to only 6 (26%), correctly ruling out the other 17 before spending any time on them. Two of those applications advanced to recruiter screens.
Tools:
- External Tools: Claude (Skills feature), Claude Design (visual summary poster)
- Internal Utilities Built: A custom verification script for automated system auditing
Methodology: I built this in two connected parts: a Project doc that holds the full, living picture of my background, voice, and job pipeline, and a Skill, a structured set of instructions Claude reads automatically, that turns that information into repeatable workflows for scoring, tailoring, and tracking. The Project doc is the source of truth; the Skill is what acts on it.
Along the way, the system caught real problems:
- A client list drifted from the truth and had to be corrected
- Six files that were supposed to exist didn't, and had to be rebuilt from scratch
- Two separate rules, a cover letter length target and a company-reputation policy, ended up saying different things in different places after being fixed in isolation, and both had to be traced back and corrected everywhere they lived
That kept happening, fix one file, find the same problem in another one later, and that pushed me to build something more solid: a script that checks the whole system for accuracy on its own, confirming the files it points to actually exist, the numbers match everywhere they should, the pay floor is stated consistently, the statuses match what is expected, and the dates are in sync, instead of relying on memory or a careful read-through to catch it. Even the tool itself needed testing. Two of its checks gave wrong answers the first time, and I only caught that by actually running it and reading the results closely.
Building this also meant staying hands-on, not just automating and walking away:
- At one point, a fix was suggested to me that I did not agree with, so I thought it through myself and pushed back, and I was right to
- Later, once new facts came in, I agreed with that same kind of fix on a different entry
- That back-and-forth, propose something, check it, correct it, check again, is really what this whole system is
Being part of AI Mastery gave me the confidence and the framework to build this properly, from the structure of the skill itself to the visual case study.
Who it's for and future plans:
Built for my own active job search, and I plan to keep using and refining it as my search continues. I'm also exploring turning it into a template others could adapt for their own job search, since the underlying system, scoring, tailoring, tracking, and self-checking, isn't specific to my situation.
Cost: $0 beyond my standard Claude subscription.
Learnings:
- A system with two places holding the same information needs a clear rule that updates get made in both places, or things will drift apart again and again, which is exactly what happened here more than once.
- I built a script that runs the check automatically, every time, because a rule I know in my head only works if something actually enforces it.
- I learned that a checker itself can be wrong, and needs to be tested against real files and read critically rather than trusted on first output.
Relevant URLs: N/A, personal tool, not publicly hosted. Full skill package and visual summary poster attached.
Attachments Included:
- Full skill zip (SKILL.md, all reference files, the checking tool, README, and the 23-entry tracker)
- Career Assistant Skill poster (Claude Design)
Written 5 August 2026·Submitted by the storyteller