Lead Compass: Scoring B2B Leads by Their Website's Health
A slow, insecure, outdated website is a warm sales signal. Lead Compass discovers prospects, scores them on deep technical signals — SSL grades, Core Web Vitals, security posture — and auto-builds a personalized demo site as the opener.
Project status: In development — discovery, scoring, and agent integration complete.
Firmographics tell you who; technical signals tell you when
Traditional B2B lead tools rank prospects by industry, size, and revenue — data that tells you who might buy, but nothing about whether they need you now. Lead Compass takes a different angle: it reads the prospect's actual website like an engineer would, because for anyone selling digital services, a struggling website is the buying signal.
The scoring engine aggregates genuinely deep signals: SSL Labs grade, Mozilla Observatory security score, DNS and WHOIS analysis, real-user Core Web Vitals from the Chrome UX Report, content freshness, schema.org validity, even the site's carbon footprint — capped by an AI scoring pass that weighs it all. A lead with an expiring certificate, failing vitals, and three-year-old content ranks above a healthy one every time, because that's the company that needs help.
An autonomous agent does the legwork
Discovery runs on a self-hosted SearXNG meta-search instance — no paid data APIs — with discovery runs monitored live from the dashboard. From there, an integrated autonomous AI agent handles deep website analysis, enhanced scoring, and the platform's most distinctive move: automatically generating a tailored demo site for a prospect, so the first outreach email contains a working preview of what their site could be instead of a pitch.
The pipeline closes the full loop: discover, analyze, score, scrape contacts, generate the demo, reach out. Embeddings stored in pgvector add semantic similarity across the lead base — find more companies that look like your best ones.
- ● Seven-service architecture — Django, Celery worker, and beat scheduler run from one shared image dispatched by environment, alongside a FastAPI crawler microservice, PostgreSQL with pgvector, Redis, and SearXNG.
- ● Real auditing tools — Lighthouse and Wappalyzer run as part of enrichment — the same tools an engineer would open manually, automated at scale.
- ● Demo-first outreach — show, don't tell: the cold email links to a working demo built for that specific prospect.
The pattern behind it
Like our other AI products, Lead Compass keeps the model on a leash: deterministic technical measurements form the backbone of every score, and the AI layer interprets rather than invents. When the number says the site scores an F on security, that's a measurement, not an opinion — and that's what makes the outreach credible.
Technology Stack
More from the build floor
WARFORGE
Mobile RTS games have always been crippled by touch controls. Our answer: an AI adjutant that executes spoken orders — parsed by a local LLM, validated by deterministic code, running entirely on your phone.
GamesSTRATUM
An archaeology game where every dig site is a procedurally generated civilization with its own history, artifacts, and a hidden mystery — guaranteed solvable by a code validator, never by trusting the AI.
GamesSymbiont
A roguelike where you grow a living organism on a hex grid, and every cell's power depends on what surrounds it. Built with zero external assets — every visual drawn in code, every sound synthesized at runtime.