😍 #promptandcircumstance #bubble.io #claude.ai

Category: Dashboard

Live Link: https://gblackman-27569.bubbleapps.io/

Sniper: I've been recruiting for 22 years. I built Sniper because every sourcing tool on the market is broken.

One job description in. Eight ranked candidates out. With outreach already written.

Sniper is a recruiter-grade sourcing workspace. Paste a job description (formal req or one sentence in plain English) and Sniper runs a seven-step AI pipeline across OpenAI, Anthropic, and SignalHire in roughly 20 seconds. The output is the top eight candidates ranked against a custom rubric, with personalized outreach drafted in your voice and live verified contact info ready to send. Every sourcing tool either gives you raw search results (Juicebox, hireEZ) or generic AI fit scores with no recruiter logic underneath (everyone else). Sniper does both, plus the part nobody does well: the outreach that follows the grade.

What it does: Seven AI calls. Three live production API integrations. One click.

1. OpenAI extracts the role title and location from any input format

2. OpenAI generates a search Boolean tuned to the actual job

3. SignalHire queries its 850 million live profile index

4. OpenAI grades each surfaced candidate against a five-pillar rubric (career trajectory, ecosystem fit, skill match, role-level calibration, risk signals) using sequential gate logic from a filed US patent

5. Claude drafts personalized outreach in the recruiter's voice

6. OpenAI runs pattern analysis across the top eight to flag compensation and market risks

7. The full dataset commits to persistent storage for cross-session pipeline tracking

The grading architecture is the subject of a filed US provisional patent (USPTO 64/048,063, filed April 2026). It implements sequential gate logic (career field match, ecosystem DNA, zero-transferability rules, senior role exemptions) before scoring across the five pillars. This is the part that took 22 years of recruiting to design and ~seven days to build into a working AI pipeline.

The recruiter workspace:

Sniper isn't a one-shot search tool. It's a working pipeline.

Every search persists to a sidebar history. Click any past run and Sniper deep-links you back to the canvas with cached results loaded instantly. Star candidates to shortlist them. Log notes per candidate ("Left voicemail," "Emailed," "Spoke with candidate") in an inline notes utility that accumulates across sessions. Reveal verified email, phone, and LinkedIn live from SignalHire's production API. Send Outreach opens your mail client with the AI-drafted message preloaded, so you send from your own account and the candidate relationship stays yours.

The Pipeline page surfaces three filters: every run, just shortlisted candidates, and candidates you've left notes on. Cross-session memory is the difference between a demo and a tool recruiters would actually use.

Built on Bubble in under seven days:

Sniper embodies the theme of this challenge: knowing when to let the AI build, and knowing when to take the wheel.

The Bubble AI app builder generated the initial structural framework and baseline layout — two pages, hundreds of elements that became the editorial design language of the final product. By letting the AI absorb the repetitive frontend scaffolding and responsive layout work, I dedicated a full week to the parts that matter: the patent-pending sequential grading logic, three live production API integrations across OpenAI, Anthropic, and SignalHire, conditional state management for the multi-call pipeline, and the persistent database structure for the recruiter workspace.

The AI laid a beautiful foundation. Twenty-two years of recruiting domain expertise built the engine.