push to main

Showcase · The Circus · 01

Merlin

A signal engine that finds companies the moment they raise capital, scores each with an AI propensity model, and surfaces the ranked results live. Running entirely on Cloudflare.

This page is a walkthrough of that engine: what goes in, what runs, and what comes back. The scoring itself happens in the Worker described in the case study.

Cloudflare Workflows Workers AI D1
Read the case study ↗
100%
Cloudflare-native. No external infrastructure
0
external API keys required to run it
0 to 100
AI-scored ICP fit for every company found

What it does

Merlin watches one buying signal: companies that just raised capital, pulled from public SEC EDGAR Form D filings. The reason to call is already on the record.

Each company is enriched, then scored from 0 to 100 against an ICP. Workers AI writes the reasoning. The ranked rows land in D1 and render on a public dashboard served from the same Worker.

The ICP is a swappable definition. The first instance scores companies against a developer-platform and edge-infrastructure ICP. Point the same engine at a different buyer and it re-scores against the new target.

How it works

Inputs

  • A weekly cron trigger, or Run now on the dashboard.
  • Public SEC EDGAR Form D filings.
  • An ICP definition. The first one is developer platforms and edge infrastructure.

Steps

  • Discover. The cron starts a weekly run, or someone hits Run now.
  • Orchestrate. A durable Workflow runs the steps and retries on its own.
  • Enrich. Workers fan out across the companies in the run.
  • Score. Workers AI acts as the propensity model: a 0 to 100 fit score, plus the reasoning.
  • Store. The rows land in D1.
  • Show. The same Worker serves the ranked dashboard.

Outputs

  • A ranked list of companies that just filed.
  • A 0 to 100 ICP fit score on every company.
  • The written reasoning behind the score.
  • A public dashboard.

Sample output

Ranked row

The fields a row carries, from the case study. No sample company is published. The card names the fields.

Signal
SEC EDGAR Form D. The company just raised.
Score
0 to 100, fit against the active ICP.
Reasoning
A short note from Workers AI on why the score landed where it did.
ICP
Developer platform and edge infrastructure, until the definition is swapped.
Where it sits
D1, then the public dashboard, highest fit first.

Source

Written from the Merlin case study. Public code: none under github.com/joelrmartinez as of 19 Oct 2026, so there is no repository link on this page.