👻 Ghosty — Never Let a Great Conversation Ghost You

---

🏆 Winner — Kiro Spark Challenge 2026

Ghosty won at the Kiro Spark Challenge 2026 at Arizona State University — a 24-hour hackathon sponsored by AWS where teams built AI-powered solutions using Kiro's spec-driven development workflow.

Frame: Economics — The Transparency Guardrail
Challenge: Expose a hidden economic factor and turn invisible data into actionable financial insight.

---

🎬 Watch the Demo

---

The Problem

Every conference ends the same way. You meet incredible people, exchange a few words — and within 48 hours, 80% of what made those conversations valuable is gone. The follow-up never happens. The opportunity disappears.

A single recruiter connection can be worth $15K–$80K in salary negotiation. A mentor relationship can shorten a career pivot by years. But nobody tracks this value. Nobody sees it. 84% of conference connections are never followed up on.

The Solution

Ghosty is a voice-first mobile conference companion that captures the context behind every human connection before it disappears — and turns social capital into a transparent, actionable career asset.

Tap the mic after a conversation. Speak for 15 seconds. Get a structured contact card with a transparent Connection Value Score showing the real dollar value of that relationship. Draft a personalized follow-up that references your actual conversation — not a generic template.

---

✨ Features

📋 Profile & Identity

• Editable profile with bio, skills, resumes, and QR code for instant identity exchange.

🗓️ Event Battle Plan

• Pre-event intelligence: attending companies, booth numbers, recruiter names, hiring signals.
• Personalized pitch hints for each company based on your skills and career goals.

🎙️ Voice Capture → AI Extraction

• One-tap recording with real microphone via expo-av.
• OpenAI Whisper transcription → Claude structured extraction.
• Generates contact cards with name, role, company, intent tag, key details, and follow-up date.

📊 Transparent Connection Value Score

• Score from 1–10 with full visible breakdown: role seniority, company tier, career relevance, intent type, recency.
• Salary band estimate with source label.
• Score decays 10% per week without follow-up — urgency is built in.

✍️ AI-Powered Follow-Up Drafts

• Personalized LinkedIn messages, emails, and cover letters referencing your actual conversation.
• AI quick actions: "Make shorter", "More formal", "Add skill highlight" — Claude rewrites in real time.

🪄 Magic Wand — Network Intelligence

• Job Search: find contacts by role or company.
• Referral Assist: one-tap prompts like "Who can introduce me to a hiring manager?"

📱 QR Quick Add

• Flash your Ghosty code for instant identity exchange at events.

---

🚀 Run Locally

npm install
npx expo start --tunnel

Scan the QR code with Expo Go on your phone.

Optional API Keys

The app runs fully in demo mode without any keys. To enable real AI:

EXPO_PUBLIC_OPENAI_API_KEY=sk-...        # Whisper transcription
EXPO_PUBLIC_ANTHROPIC_API_KEY=sk-ant-... # Claude extraction & drafting

---

🏗️ Architecture

flowchart TD
  A[Sign In] --> B[Profile]
  B --> C[Events — Company Intel]
  C --> D[Attend Booth — Voice/Text Capture]
  D --> E[Whisper Transcription]
  E --> F[Claude Extraction]
  F --> G[Transparent Scoring]
  G --> H[Contact Card]
  H --> I[Follow-Up Queue]
  I --> J[AI Draft Editor + Quick Actions]
  B --> K[Magic Wand — Network Search]
  B --> L[QR Quick Add]

Loading

📁 Project Structure

.kiro/                 Specs, hooks, and steering docs
src/components/        Voice, contact card, follow-up, dashboard UI
src/hooks/             Recorder (expo-av), queue, alert state
src/screens/           SignIn, Profile, Events, FollowUp, Wand, QR, ContactDetail
src/services/          Whisper, Claude, scoring, calendar stub, Supabase stub, Ghosty AI
src/data/              Sample contacts (real mentors), events, profile
supabase/              Schema migration and edge function (production-ready)

---

🛠️ Built With

Layer	Technology
Frontend	React Native + Expo
Voice Capture	expo-av
Transcription	OpenAI Whisper API
AI Extraction & Drafting	Claude (Anthropic)
Scoring	Custom transparent algorithm
Backend Schema	Supabase Postgres + RLS
Development	Kiro — spec-driven development
Language	TypeScript (strict mode)

---

🧠 Built with Kiro

Ghosty was developed using Kiro's spec-driven workflow:

• 6 specs defining requirements before code
• 3 steering docs guiding AI extraction, scoring, and UI design
• 4 agent hooks documenting reactive workflows
• Vibe coding for UI iteration and feature refinement

See KIRO_USAGE.md for the full write-up on how Kiro shaped the build.

---

👥 Team

Built in 24 hours at the Kiro Spark Challenge 2026 at Arizona State University.

---

📄 License

MIT — see LICENSE