AI-First
Career Map
I used AI agents to read and tag 287 real AI job listings, then turned the results into a private tool that tells me which roles fit, which skills matter, and what to build next.
The question
Which AI jobs suit someone who builds real software with AI coding agents, but does not hand-write code?
The data
287 listings: 133 in the Orlando area and 153 remote, as of 2 Nov 2026.
Where AI is used
Agents collect and read postings, tag every requirement from a fixed list, and scan for new jobs each morning.
Built with
- Claude Code
- Subagents
- Python
- Scheduled tasks
Job titles do not say what the job is.
"AI Engineer" can mean training models, wiring APIs, or running a customer workshop. Job boards sort by title, so they could not tell me which roles value practical AI building and which quietly require years of hand-written code.
I needed evidence, not a feeling: what employers actually require, what they only prefer, and how often.
- The same title hides very different work.
- Required and "nice to have" items are mixed together in each posting.
- New listings appear daily; a one-off study goes stale in a week.
- Reading hundreds of postings by hand is slow and inconsistent.
Read every posting the same way.
The key decision was a closed tag list. Every requirement maps to one of about 120 fixed tags, so counts can be compared across hundreds of postings instead of drifting with each reader's wording.
Parallel collection
Several subagents searched local and remote listings at the same time, each with its own output file so they could not overwrite each other.
Required vs. preferred
Each posting was tagged twice: what it requires and what it only prefers, with the original wording kept as an excerpt for checking.
Plain analysis
The counting, deduplication and comparisons are a standard-library Python script. No AI in the maths, so every number can be re-run.
Self-updating
A scheduled task scans for new listings every morning, applies the same qualification rules, adds them to the data, and rebuilds the report.
The most requested skills.
Share of 287 AI listings that ask for each skill, required or preferred.
What changed because of it
Agents and evaluation are near the top, so my case studies lead with agent pipelines and measured AI quality. Communication and demos matter to more than half of employers, so I built this portfolio to explain my work, not just list it.
What I did. What the AI did.
- Framed the question and the job families worth studying.
- Set the closed tag list idea, so results stay comparable.
- Wrote the qualification rules: a job is dropped if any required item is something I cannot honestly claim.
- Decided which sources were allowed and which tools needed my approval first.
- Read the findings and changed my plan because of them.
- Ran the parallel collection and tagging agents.
- Wrote the Python analysis and report builder.
- Ran random spot-checks of tags against the original postings.
- Runs the daily scan and rebuilds the report each morning.
- 29 Oct 2026 · Research
First study and report
Listings collected and tagged, analysis written, report built.
- 30 Oct 2026 · Automation
Daily scan with a qualification gate
New listings are checked against the required-items rule before they can appear.
- 1 Nov 2026 · Iteration
New jobs feed the skill counts
Every new listing read in full is tagged the same way, so the skill rankings keep moving with the market.
- 2 Nov 2026 · Integration
Evidence map and interview mode
The map now links each skill to the project that proves it, and builds interview presentations for a specific job.
Public method. Private data.
This page shows the method and the market-level findings. The tool itself runs only on my own computer.
- Not published: which jobs I applied to, or my notes on any employer.
- Not published: my fit scores, gaps and interview preparation.
- Published: aggregate counts across all 287 listings, and how they were produced.
The same method fits other messy decisions.
Vendor or tool selection
Read every proposal or product page against one fixed checklist, then compare like for like.
Market and competitor research
Tag competitor offers, pricing and claims the same way every week, and see what changes.
Support and feedback analysis
Turn free-text tickets or reviews into counted, comparable themes with the original wording kept for checking.
Daily monitoring
A scheduled agent that checks a source each morning, applies your rules, and updates one report.
Let's talk about
what you need built.
Open to AI roles: full-time, contract or consulting. I build AI products, internal tools and automations with AI coding agents.