// Project case study
Neurix Intelligence
An internal lead-intelligence platform that finds businesses across North America and Europe with no website — or a broken, outdated one — proves it with stored evidence, enriches public contact details and flags regional outreach rules.
- Role
- Architect & developer
- Duration
- Nov 2026
- Stack
- PythonSupabase / PostgreSQLVercelRenderDockerSearXNGOverture Maps
The Problem
Lead lists are stale, unverifiable and expensive — and outreach rules differ by region.
The Approach
A queue-driven pipeline: cloud web app and API, Supabase job queue, and a local worker that discovers and verifies businesses with stored evidence.
The Outcome
Evidence-backed leads across North America and 46 European countries, exportable to Excel, CSV and PDF, on free infrastructure.
Internal tool for Neurix Studio.
Who it's for
The Neurix Studio growth team: people who need a reliable, evidence-backed list of businesses that genuinely need a web presence, without paying for lead databases.
The challenge
Lead lists are usually stale, unverifiable and expensive. The team needed to search whole countries, check each business's real web presence, keep proof of every claim, and stay within regional outreach rules — on free data and free infrastructure.
What I built
A queue-driven system: a Vercel web app and a Render API sit on Supabase (PostgreSQL), while a worker on a local machine claims jobs, searches Overture Maps and OpenStreetMap, checks each business's website and search presence through a private SearXNG instance, and writes evidence, contacts and scores back to the database. Results flow into a CRM, outreach drafts and Excel, CSV or PDF exports.
Why it matters
It turns lead generation from guesswork into a verifiable pipeline. Every lead carries the evidence for why it qualifies, and the architecture keeps hosting at zero cost by moving the memory-heavy search work to a worker instead of the cloud tier.
How it works
- 01Pick a country, an area (state, county, city, bounding box or radius) and one or more niches, then estimate the run size.
- 02The API queues the job in Supabase; a worker claims it and discovers businesses from Overture Maps (optionally OpenStreetMap), dropping chains and closed places.
- 03Each business is verified one by one — website status, search presence and public contact channels — with evidence stored alongside the lead.
- 04Leads land in the dashboard and CRM with regional outreach notes, and can be exported to Excel, CSV or PDF.
Scope note
Neurix Intelligence is an internal tool, so there is no public demo or repository. This case study describes its architecture and workflow, not commercial results.
System Architecture
- 01
Web app
Vercel
- 02
API
Render · auth · CRM
- 03
Queue & data
Supabase · PostgreSQL
- 04
Worker
Overture Maps · SearXNG
- Five-part architecture: web app (Vercel), API (Render), database and job queue (Supabase/PostgreSQL), worker, and SearXNG in Docker.
- Single-claim job queue so multiple workers never process the same run twice; runs are resumable.
- Covers the United States, Canada, Mexico and 46 European countries; a separate startup pipeline tracks new UK companies, SEC Form D filings and product launches.
- Runs entirely on free data sources and free-tier infrastructure — no paid API keys.
// Learnings
What this project taught me
- 01A job queue in the database is often all the infrastructure you need.
- 02Store the evidence with the claim — it makes every lead defensible.
- 03Move memory-heavy work to where the memory is.
// Related work
Next project
CampusLoop