Forex Lead Finder.

Manual lead hunting burns a full workday per week.

Target buyer: anyone selling forex signals, courses, or brokerage services. Their bottleneck: finding qualified traders — people with MT4/MT5 bios, trading-pair posts, or telltale hashtags. That audience is scattered across Reddit, TradingView, BabyPips, ForexFactory, and LinkedIn. Existing tools: cover one platform at a time, require constant maintenance, and return unscored data that still needs a human to filter.

Five scrapers in parallel, scored at ingestion, sales-ready on export.

What ships: five platform-specific Python collectors running concurrently; a 15-point keyword classifier gating every candidate (bio keyword +10, post keyword +10, trading pair +10, MT4/MT5 +5, hashtag +5, threshold 15); SQLite warehouse keyed by profile URL (idempotent re-runs, no duplicates); Flask UI for operator review; REST endpoints for downstream CRM / email tools; APScheduler 6-hour loop; XLSX / CSV / JSON export.

What it replaces: a junior researcher spending ~10 hrs/week scraping names manually (typical market cost in Dhaka: $800–1,200/month; US: $3,500–5,000/month). This engine compresses that to zero human time after initial tuning.

Non-obvious decisions.

• Rate-limit isolation — each platform has its own throttle config. A spike on one source never triggers a ban on another.
• SPA vs static detection — Playwright only boots where the platform requires JS render; the rest run through requests + BS4 for 10× speed.
• Classifier calibration — the 15-point threshold was tuned against a labelled seed set; accepts roughly 26% of scraped profiles and rejects 74%.

Why this one matters.

A single operator ships a Python system that replaces the work of two or three manual lead researchers. The product proves the hybrid-SEO-plus-engineering positioning: commercial keyword knowledge (what makes a trader "qualified") fused with the Python chops to automate the capture.