ScoutWise

Your AI agent for smarter, faster home hunting

Comment

Inspiration

Finding a rental in a major city can feel like a full-time job — especially during a stressful relocation. I experienced this firsthand during my move to Munich. Between comparing dozens of listings, parsing vague descriptions, evaluating photos, and chasing down unresponsive listers, I found the process exhausting and inefficient. That experience inspired me to build ScoutWise — an AI agent that could take the burden off the renter and become a proactive assistant in the house-hunting process.

What It Does

ScoutWise is an intelligent house-hunting agent that:

  • Takes in your preferences (location, price, number of rooms, desired amenities, and style)
  • Automatically scrapes fresh rental listings from Zillow via Apify
  • Uses Clarifai’s AI models to analyze the first two images of each listing to evaluate visual quality and style fit
  • Summarizes and ranks listings based on both text and image data
  • Generates and optionally sends a tailored inquiry email to the lister on your behalf

It’s designed to not just recommend — but act.

How We Built It

  • Frontend: React + Tailwind, with a clean user form to collect preferences
  • Backend: Node.js + Express for the agent logic and orchestration
  • Web Scraping: Apify’s Zillow scrapers (ZIP search + detail scraper) to retrieve and enrich listing data
  • AI Analysis: Clarifai’s gemma-3-12b-it multimodal model to evaluate listing images based on user-stated style preferences
  • Email Drafting: Agent generates pre-filled inquiry drafts to help the user reach out efficiently

Challenges We Ran Into

  • Understanding and integrating multiple Apify actors in a chained workflow
  • Parsing and filtering inconsistent listing data (some had no images, others lacked contact info)
  • Making Clarifai’s image analysis meaningful in a subjective domain like “style”
  • Structuring the agent logic to feel autonomous without overcomplicating the MVP

Accomplishments We’re Proud Of

  • End-to-end working pipeline: from user input → web scraping → AI evaluation → actionable output
  • Image analysis tailored to user-described style preferences using Clarifai’s LLM-based vision model
  • Clear separation of agent components (scraping, reasoning, acting) while keeping the user flow simple
  • A frontend that feels user-friendly despite the complexity under the hood

What We Learned

  • How to use Clarifai’s gRPC API to query multimodal AI models
  • How to orchestrate multi-step agents that chain tools (Apify + Clarifai + custom logic)
  • The importance of balancing automation with user control in agent UX
  • How real-world pain points (like housing) make the best use cases for agent-based tools

What’s Next for ScoutWise

  • Enable recurring background runs via n8n to monitor listings and auto-reach out in real-time
  • Integrate more listing platforms beyond Zillow
  • Use natural language summarization to auto-generate pros/cons for each listing
  • Build in feedback loops where the user can mark “like/dislike” to improve agent matching
  • Add SMS/email notifications for top listings when they appear
Share this project:

Updates