Back to projects

Deep Research

2024shipped

Turns an email address into a buyer dossier using Gemini's async research API for luxury real estate lead qualification.

// GitHub

View Repository
33 commits
Last commit 2 months ago
TypeScript

// Problem

Luxury real estate agents can't distinguish billionaires from tire-kickers when leads come through a website form. Without knowing a prospect's actual buying power, high-value opportunities get the same generic follow-up as everyone else.

// Solution

When a lead shows interest, Gemini's Interactions API runs background research and synthesizes a dossier with net worth signals, professional history, and a clear verdict on whether they're worth pursuing. Agents see the full picture before picking up the phone.

// What I Built

Full-stack platform with a public listing site and agent dashboard. Property likes trigger async Deep Research jobs that backfill structured data into lead profiles. Agents see leads grouped by status with dossier panels showing assets, social signals, and engagement recommendations. A second-pass SSE enrichment system extracts structured fields with confidence scores for agent review.

// Technologies

Google Gemini Interactions API

Async research jobs that return structured intelligence reports without blocking the UI.

Next.js 16 + React Server Components

Server actions handle all mutations; sensitive logic stays server-side.

PostgreSQL + Drizzle ORM

JSONB columns for flexible enrichment data that can evolve without migrations.

Server-Sent Events (SSE)

Streams enrichment progress and field extractions to the UI in real time.

GCP Service Account OAuth

Service account auth with bearer tokens instead of API keys.

// Lessons Learned

  • 01Async AI jobs fit research workflows perfectly. Kick off a job, poll for results, display when ready.
  • 02OSINT techniques translate directly to sales intelligence. Triangulating identity and assessing credibility is exactly what lead qualification needs.
  • 03SSE streaming beats waiting for full responses. Progressive UI hydration with confidence scores creates a better experience.
  • 04Structured output via responseSchema eliminates parsing bugs. Guaranteed JSON from Gemini is a reliability game-changer.