RolesRadar: how I built an AI job search agent on n8n
RolesRadar is an AI job search automation I built on n8n. It reads my resume, scans live listings daily, and ranks the best-fit product and AI roles for me.
I got tired of scrolling job boards, so I built RolesRadar: an AI job search agent that runs every day without me. It reads my resume, understands what I actually want, searches live listings across the web, and hands me a short list of the roles worth applying to.
The whole thing is an n8n workflow orchestrating two Gemini calls, the Google Custom Search API, Google Sheets, and Gmail. No servers to babysit, no scraping, no brittle glue code. Just a schedule, a pipeline, and a daily digest in my inbox.
Step one is a schedule trigger that fires at noon every day. Step two pulls my resume text and passes it to a Gemini prompt I call the Career Direction Analyzer. That prompt turns my raw resume into a structured JSON profile: target titles, target domains, keywords, exclusions, and an experience band. This is the single source of truth for the rest of the run.
Step three is a small code node that builds a boolean search query from that profile and strips out staffing agencies, contractor marketplaces, and other noise. Step four sends that query to the Google Custom Search API and returns fresh job listings from real company sites.
Step five is where the second Gemini call earns its keep. It scores each listing 0 to 100 against my profile, drops anything senior or 5+ years, and returns the top 10 matches with plain-English reasons for the score. Step six flattens and appends those matches to a Google Sheet, and step seven sends a Gmail digest so I know the fresh list is ready.
The interesting part was not the tooling. It was treating the job hunt like a product. I mapped the underlying jobs to be done (find fit, filter fast, avoid burnout), wrote acceptance criteria, and only then picked the stack. Prompt engineering became spec writing. The workflow became the product.
If you are early career in product, AI, or program management, this kind of automation is a cheat code. It compounds. Every day the agent runs, I get sharper signal and less noise, and I get to spend my time on tailored applications instead of tab overload.
RolesRadar is live and running. If you want to see the full case study, workflow diagram, and the exact pipeline steps, head to the RolesRadar project page on this site.