Job Hunter

One-liner: A personal job-hunting dashboard that scrapes five job boards, AI-ranks every listing, benchmarks the pay against government wage data, and pings me on Discord when a good one lands — the Build Loop pointed straight at my own career.

Origin (the Build Loop)

Because I was actively pursuing operations/AGM roles, I kept hitting the same drowning-in-browser-tabs problem — the same listings across five sites, no way to tell a real offer from a lowball, and good roles gone before I refreshed, so I built a dashboard that scrapes the boards, scores each job with Claude, benchmarks salary against BLS data, and alerts me on Discord, that turns a chaotic search into a ranked, tracked pipeline. It taught me that the strongest proof of “I build the tools for the job I’m doing” is building the tool for the job search itself.

The problem

Job hunting for the exact kind of role this portfolio is about — ops/AGM — meant the same friction every day: the same listings duplicated across LinkedIn, Indeed, Glassdoor, ZipRecruiter, and Google; no fast way to tell whether a posting was worth my time or a lowball; and the good ones filled before I got around to checking. I wanted a system that did the watching and the first-pass judgment for me.

What it does

• Multi-board scraping — LinkedIn, Indeed, Glassdoor, ZipRecruiter, and Google via python-jobspy.
• AI ranking — scores each job across pay, flexibility, location, requirements, hours, and workload using Claude.
• BLS wage benchmarks — compares each salary against Bureau of Labor Statistics Indiana wage data, so “is this pay real?” gets an answer.
• Salary targeting — set minimum / target / stretch salaries and benchmark against the data.
• Discord alerts — scheduled scrapes ping a channel the moment a job clears my score threshold.
• Tracking pipeline — import jobs into a tracker with status, tier, notes, and resume tailoring, backed by a stored master resume.
• Market Analysis dashboard — visual breakdown of every scraped job: sectors, salary distribution, remote vs. onsite, top companies, posting timeline.
• Dedup everywhere — cross-run deduplication (never analyze/alert the same listing twice) and import duplicate protection.

How it’s built

• Stack: Next.js 15 App Router, React, Tailwind, shadcn/ui; SQLite via better-sqlite3; Anthropic Claude API; python-jobspy driven as a Python subprocess; recharts for the analysis dashboard; papaparse; Discord webhooks.
• Notable engineering:
  • Node/Next front end orchestrating a Python scraping subprocess — a clean polyglot bridge rather than reimplementing the scraper.
  • AI scoring across six weighted dimensions, not a single “good/bad” flag.
  • Cross-run dedup and scheduled scraping make it a standing service, not a one-shot script.
  • Real external data integration (BLS wage benchmarks) to ground the salary judgment.
• Architecture: scheduled scrape → python-jobspy subprocess → dedup → Claude scoring + BLS benchmark → SQLite → dashboard + Discord alert → tracking pipeline.

Proof points

• 5 job boards aggregated through one pipeline.
• 6-dimension AI scoring per listing.
• Government wage-data benchmarking, not just scraped salary text.
• Deduplicated, scheduled, alerting — a real running system.

What to show

• Demo: Deploy to a public URL (repo is already public). Seed with a live scrape so a visitor sees the ranked list, the Market Analysis charts, and a sample Discord alert.
• Visuals needed: the ranked job list with scores; the Market Analysis dashboard (sector/salary charts); a Discord alert screenshot.

Cross-links

• Shares the scrape → AI-score → deliver shape with Hermes and the “AI says is-this-worth-it” judgment in thrift-lens and POLYBOT.
• It’s the living proof behind the Where It’s Heading pillar.

Case-study angle

The single most on-thesis project in the portfolio: Alex built his own AI-ranked job-search platform to run the exact search these applications are part of. If you want to know whether he actually turns friction into systems — he did it to find you.