Fred McCullough ยท AI Automation Engineer

Multi-agent systems for messy operational data.

I build working tools for semantic search, browser automation, lead intelligence, audit pipelines, and operational decision support.

Start with the flagship case study Verify on GitHub Run CloudScan Contact

Portfolio Fast-Path

LeadOps: 8,406 public records

A local business corpus cleaned, structured, and exported into public-safe semantic artifacts that power the flagship explorer.

Read the case study

Semantic visualization

A Three.js and Leaflet interface for searching, walking, and mapping semantic neighborhoods without flattening the work into a table.

Open explorer demo

Agent-assisted delivery

Scoped AI worker lanes, archived logs, taskboard evidence, and peer review make broad automation work reviewable as implementation steps.

View GitHub profile
Featured work

Systems that make dense data useful.

The portfolio leads with production-grade interactive tools, separating polished case studies from internal utilities and older experiments.

Flagship case study

Montgomery County Semantic Explorer

A live browser interface over 8,406 local business records: semantic search, graph navigation, neighborhood walking, map handoff, and mobile-hardened interaction states.

Read the case study Open the app
semantic search Three.js Leaflet 8,406 records
Website audit tool

CloudScan

Performance, security, SEO, accessibility, and technology signals wrapped as a fast public scanner.

Run a scan
Local web intelligence

The Pulse

A county-level digital presence snapshot showing visibility, contactability, security, and speed trends.

Open the snapshot
How the work connects

One operating model, several surfaces.

The experiments share the same pattern: gather messy public signals, make them inspectable, then turn them into tools a person can actually use.

Data intake Public records, website signals, business profiles, and local-market observations.
Verification Deterministic checks, visible evidence, QA passes, and browser-level review before promotion.
Interaction Dense data becomes searchable, navigable, and usable through focused browser interfaces.
Delivery Client work, demos, and public case studies are separated from drafts and internal staging.
Supporting surfaces

Useful, but not the front door.

Some tools are operational prototypes or historical experiments. They remain reachable when useful, but the public story starts with the strongest finished artifact.

Client operations prototypes

Lead-review tools stay private until their public/private data boundary is explicitly cleaned.

Older explorer builds

Historical semantic-demo variants should resolve toward the current polished explorer.

Current app