Work / Client Lead Crawler
Leads that find you first.
A 24/7 lead-intelligence system on an NVIDIA Jetson Nano that scans public records across metro Memphis every morning, scores each lead for urgency and alerts the sales team in Slack.
Focus
Automated lead intelligence
Hardware
NVIDIA Jetson Nano, on site
Pipeline
Python · NER · geo-spatial scoring
Output
Google Sheets · Slack alerts
The problem
Good leads, found too late.
The sales team spent hours researching prospects by hand: scanning the news, cold calling and waiting on referrals. That reactive approach kept missing the best opportunities:
- Businesses that had just had a break-in
- New commercial buildings under construction
- Newly licensed businesses with no security in place
They needed those leads surfaced automatically, before competitors could act.
The solution
How it works.
Every morning at 8:00 AM, collectors pull four public data sources. The pipeline on the Jetson Nano cleans up the records, recognizes businesses and addresses, maps them and scores each lead from 0 to 100. The full list lands in Google Sheets, and anything scoring 70 or more goes straight to Slack.
Data sources
Four public sources, every morning.
Free, public and collected automatically at 8:00 AM. No one has to go looking.
| Source | What it finds | Coverage |
|---|---|---|
| Memphis Police Department | Commercial burglaries, robberies, theft, vandalism | City of Memphis, updated daily |
| Shelby County building permits | New commercial construction and renovation | Shelby County, TN, continuously |
| Memphis business licenses | Newly registered commercial businesses | City of Memphis, continuously |
| SAM.gov federal contracts | Government facilities needing security services | Tennessee, Mississippi and Arkansas |
Scoring
Every lead gets a score.
Scores run from 0 to 100, weighted by crime severity, repeat targeting, crime density within 5 miles, property value, active construction and business type.
| Priority | Score | What happens |
|---|---|---|
| High | 70–100 | Immediate outreach, with a Slack alert in seconds |
| Medium | 40–69 | Added to the active pipeline |
| Low | 0–39 | Monitored |
Key capabilities
The pipeline, step by step.
01
Data ingestion
Collectors run each morning against four live sources, handling rate limits, paging and format quirks.
02
Normalization
CSV, JSON, HTML and PDF records become one schema, with addresses standardized for matching.
03
Entity recognition
NER picks out business names, owners and incidents, and links the same business across sources.
04
Geo-spatial analysis
Every record is mapped across 11 counties, with crime density at 0.5, 1 and 5 miles.
05
Scoring engine
A weighted, multi-factor model scores each lead from 0 to 100.
06
Dedup and enrichment
Duplicates merge and compound their scores; owner contacts come from public assessor records.
07
Alerting
Leads scoring 70 or more reach Slack within seconds; a daily digest updates Google Sheets.
08
Continuous learning
Won and lost deals feed back into the weights to sharpen future scores.
Results
At a glance.
24/7
automated, no manual input
4
public data sources scanned daily
11
counties across Tennessee, Mississippi and Arkansas
$0
to run, beyond the hardware already in place
Coverage: Shelby, Tipton and Fayette counties in Tennessee; DeSoto, Marshall, Tate, Coahoma, Tunica and Benton in Mississippi; Crittenden and St. Francis in Arkansas. All data travels over an encrypted private VPN.
Next project
Solar Fleet Dashboard
Health, battery risk and location for every remote solar site, on one screen.
