Personal project

Opportunity Radar

Surfacing company hiring momentum before it's public

Project Overview

Opportunity Radar

Self-directed

Opportunity Radar is a live dashboard I built to track company hiring momentum in real time. It scans funding announcements, executive hires, news, and community signals across seven independent sources, then scores and ranks companies by how likely they are to be hiring soon — surfacing opportunity before it hits a job board.

Project Tools

n8n

Serper API

Claude

Figma

VS Code

Project Skills

Workflow automation

API integration

Signal scoring

Prompt engineering

UI design

Full-stack build

Project Team

Kevin

Claude

Problem Statement

Job seekers waste hours scrolling job boards, reacting to postings that already have hundreds of applicants by the time they appear.

The real signal — a company about to hire — shows up weeks earlier, scattered across funding news, executive moves, and community chatter.

Challenges, Opportunities & Goals

Challenge: Hiring signals that predict a job opening are scattered across dozens of sources — no single tool aggregates or scores them.

Opportunity: Build an automated pipeline that scans multiple leading indicators — funding, exec hires, hiring activity — and ranks companies by real hiring momentum.

Goal: Ship a working, self-hosted tool that scores companies daily and surfaces the strongest opportunities before they're publicly posted.

Building the Pipeline

Company Discovery (Serper + GPT)

An n8n workflow searches for newly funded and newsworthy companies, then uses GPT to extract clean company names from the raw search results.

  • Two parallel Serper searches scan for company launches and funding news.
  • GPT extracts and validates real company names from noisy search results.
  • Discovered companies merge with a static tracking list into one canonical dataset.

I started in n8n to build the internal mechanism to make it all work...

Signal Aggregation & Scoring

A weighted formula normalizes and combines every signal into a single opportunity_index, capping each source so no single signal can dominate the score.

  • Weighting: Funding signals carry the highest weight as the strongest leading indicator.
  • Normalization: Each signal is capped, then scaled 0–1 before being weighted.
  • Output: A single 0–1 opportunity score, recalculated on every run.

Weighting & Signal Design

Each signal source was tuned individually — querying, parsing, and weighting it to reflect how strongly it predicts near-term hiring.

  • Funding signal: Scans TechCrunch, Crunchbase, and PitchBook for recent raises.
  • Exec-hire signal: Scans LinkedIn for senior 'started a new position' announcements.
  • Scoring formula: Capped, normalized signals combined into one weighted 0–1 index.

Building the Tool

Figma UI Design

I designed the dashboard layout in Figma — a signal radar visualization and a ranked company list — then exported the design as JSON and handed it to Claude to generate the front-end build.

  • Figma design exported as structured JSON, not just visual reference.
  • Claude generated the initial HTML/CSS build from the design JSON.
  • Iterated in VS Code, going back and forth on layout and logic.
  • Matched the final build to the original design intent.

With the backend scoring pipeline working, I moved into building the interface — pairing a Figma design with Claude-generated front-end code, refined through iteration in VS Code.

Live Dashboard

The finished dashboard shows a signal radar view, a ranked list of tracked companies, and a Company Inspector panel explaining why each company is trending.

  • Signal radar visualizes momentum — distance from center reflects score strength.
  • Company Inspector breaks down why a company is trending, source by source.
  • Ranked list auto-updates with each new pipeline run.
  • Live and publicly viewable, hosted on GitHub Pages.

    Live Tool

    Outcomes

    Opportunity Radar went from concept to a live, working tool by pairing automated backend workflows with an AI-assisted front-end build — a full pipeline shipped solo, end to end.

    7

    Independent signal sources tracked

    14+

    Companies scored per run

    1

    Person build, powered by Claude

    Shipped a fully working dashboard, live and publicly accessible.

    Proved a weighted, multi-source scoring model can meaningfully differentiate companies.

    Built the entire pipeline — backend, scoring, and UI — using Claude as an AI pair-programmer throughout.

    Key Takeaways & Lessons Learned

    Building a real signal pipeline meant debugging data end-to-end — from a single wrong field name silently breaking a formula, to signals that looked strong but weren't actually differentiating anything.

    A single mismatched field name can silently break an entire scoring formula — always verify output, not just that a node ran.

    A signal that's maxed out for every company isn't actually a signal — it needs a tighter query, not just a slot in the formula.

    Pairing with Claude across n8n, Figma, and VS Code made it possible to move from idea to shipped tool in one continuous build session.

    Project Assets

    A look at the live tool and the workflow behind it.

    Copyright 2025 Kevin Keiper