JobRadar — AI Freelance-Order Radar
A local desktop app that pulls freelance orders and jobs from a dozen platforms, scores each one against my profile with AI, and surfaces the best matches in one dashboard.
Problem :
As a freelance developer, the good orders are scattered across a dozen job boards and Telegram channels, and the best ones disappear within hours. Checking every source by hand is a part-time job on its own — and you still miss the ones that actually fit.
Solution :
JobRadar collects listings from many sources into one local database, scores every order with AI against my profile — is it real dev work, does it fit, what’s the budget, how reliable is it — and surfaces the top matches in a dashboard, with Telegram alerts for the strongest ones.
My Role :
Everything — the product idea, architecture, UI design, the collectors, the AI scoring pipeline, the dashboard and the desktop packaging.
Process :
Built as independent modules: collectors that scrape each source, a scoring pipeline that ranks every order, a notifier that pings Telegram, and a Next.js dashboard to triage the incoming feed, run a Kanban of applications and read stats. It runs fully local — its own SQLite database and server — scheduled on Windows and packaged as an Electron desktop app.
Outcome :
One place that turns a firehose of listings into a ranked, filterable feed of orders that genuinely match — each with a match score, a reliability flag and a budget. It’s my own daily tool; the screens here are live data.
Tech Stack :
Next.js + TypeScript, Drizzle ORM over SQLite, Electron for the desktop app, per-source collectors (incl. Telegram via GramJS), AI scoring through the Anthropic API, Telegram notifications, scheduled with Windows Task Scheduler.
