lead-gen ai product
The request came from management and sales: they needed an AI tool that brings in leads and gives the team a reason to start a conversation. Free for the user, cheap for the company to run. We chose to analyse venues from their Yandex Maps reviews — that way we reach the ones who aren’t on the platform yet.
I built the whole thing, from funnel logic to deploy. An owner sends the venue name or a link and gets back a pdf audit of their own reviews. Effectively a lead generator with no ad budget.
how it works
The user sends a venue name or a link. The scraper collects reviews from Yandex Maps, and a model sorts them into categories on a custom prompt and scores the venue.
Out comes a pdf: how the venue looks online, where it slips, what to fix first, and which paid tools of ours would help. The file is assembled in code — brand fonts, colours, logos, layout.
how a lead reaches sales
After the report the bot shows a personal offer. One click and a card lands with a manager: who asked, when, and about what, plus the same pdf. From there they sell with an analysis in hand instead of a cold call.
under the hood
- python, aiogram on long polling, fastapi for the admin panel
- postgres with alembic migrations, redis for the task queue
- two scrapers: apify — fast and easy on the queue, but paid by volume; and my own selenium with undetected-chromedriver — the free fallback if the first one goes down
- analysis on a custom prompt: categories and a score for the venue
- pdf built in fpdf2, matplotlib and pillow — brand layout written in code
- admin panel: requests, stats, handing a lead to a manager
- deployed in docker, migrations applied by script
results