← Back to portfolio
Case Study · Automation · Solo Build

AI Automation Systems

Scan, score, route. Bots that watch external sources, judge what matters, and act on a schedule — without anyone touching them.

Role
Sole designer & builder
Pattern
Scan · Score · Route
Status
Running on cron
Delivery
Email · Telegram
01

The brief I set myself

Take a repetitive watch-and-react job a person does by hand — checking a source, judging each item, then acting on the good ones — and turn it into a system that runs itself. The first build hunts freelance opportunities: it scans listings, scores each against my criteria, drafts a tailored proposal, and sends me a shortlist. Same pattern feeds any pipeline — leads, tenders, mentions, price changes.

02

The pattern

01 · Scan

Pull

Watch external sources on a schedule — job boards, APIs, sites — and collect new items into one clean feed.

02 · Score

Judge

Rank each item against custom rules and an AI judgement pass, so only what actually matters rises to the top.

03 · Route

Act

Draft the response, create the task, and push the alert out — to email or Telegram — ready to action.

Unattended

Runs on cron, not on me

Scheduled jobs with logging and error handling. It keeps working after handover — the whole point of automation is that you stop thinking about it.

Judgement, not just filters

Scored against real criteria

Rules plus an AI scoring pass mean it surfaces the genuinely good matches and drafts a tailored reply — not a dumb keyword dump.

03

What it demonstrates

Reliable automation — scheduled, supervised jobs that run in production without babysitting. A reusable pattern — scan/score/route maps onto lead capture, tender-watching, monitoring and follow-ups. AI where it earns its place — scoring and drafting, wired into a system that actually acts.

04

Built with

Python Anthropic Claude Scheduled jobs (cron) Web scraping / APIs Scoring engine Telegram Bot API