Helio — AI-Powered Real Estate CRM

Inspiration

Finding accommodation is a recurring challenge, especially for students and first-time renters in competitive markets like Los Angeles. Each cycle, students face tight timelines, fragmented listings, and limited transparency around pricing, risk, and neighborhood quality.

Most platforms optimize for browsing, not decision-making. Users are forced to piece together information across multiple tools, maps, listings, notes, and external research, while under time pressure. This often leads to rushed or suboptimal housing decisions.

Helio was inspired by this gap. We set out to build a system that does not just show properties, but helps users evaluate, compare, and decide, within a single, structured workflow.


What We Built

Helio is an AI-powered real estate CRM that unifies search, analysis, communication, and visualization into one platform.

At its core is the Deal Score, a weighted model that evaluates each property across five dimensions:

$$ \text{Deal Score} = 0.30V + 0.25L + 0.20C + 0.15M + 0.10R $$

Where:

  • (V) = Value (price relative to comparable properties)
  • (L) = Location (neighborhood quality, accessibility, and growth signals)
  • (C) = Condition (renovation needs and hidden costs)
  • (M) = Market Momentum (demand and pricing trends)
  • (R) = Risk (downside exposure, environmental and financial risk)

This transforms raw listings into structured, explainable insights that support faster and more confident decisions.

Helio integrates multiple capabilities into a single system:

  • Property search with live listings and map-based visualization
  • AI-powered comparison that selects the best property based on user preferences
  • A context-aware AI assistant with both text and voice interaction
  • 3D property tours using procedural generation and embedded models
  • AI-generated floor plans and cinematic walkthrough videos
  • Downloadable reports and email summaries for sharing decisions

How We Built It

Helio is implemented as a full-stack web application with a modular, API-driven architecture

  • Next.js 16 and TypeScript for the application framework
  • Tailwind CSS and shadcn/ui for UI development
  • Mapbox for geospatial visualization
  • Gemini 2.0 Flash and OpenAI GPT-4o for AI reasoning and generation
  • OpenAI image models and Sketchfab for floor plans and 3D tours
  • Three.js and React Three Fiber for interactive 3D environments
  • ElevenLabs and Groq Whisper for real-time voice interaction
  • Cloudinary for image storage and CDN delivery

The system is designed with fallback behavior, allowing core functionality to operate even without external API dependencies.


Challenges

Integrating multiple AI modalities
Combining text, voice, image generation, and video into a single coherent workflow required careful orchestration of APIs and latency management.

Designing a reliable scoring system
The Deal Score needed to be both interpretable and meaningful. We focused on transparency and clear breakdowns to ensure users can trust the output.

Maintaining real-time performance
Features like map interaction, AI responses, and 3D rendering had to remain responsive despite heavy computation and external API calls.

Unifying complex workflows
Real estate tools are typically fragmented. Bringing search, analysis, communication, and visualization into one interface required consistent data models and state management.


What We Learned

  • Users struggle more with decision-making than with access to listings
  • Explainability is critical for trust in AI-assisted systems
  • Combining structured data with AI produces significantly better outcomes than either alone
  • Multimodal interfaces (voice, visual, spatial) can reduce friction in complex workflows

What’s Next

  • Expanded live listing integrations
  • Persistent user accounts and collaboration features
  • Agent-client shared workspaces
  • Transaction and offer generation workflows

Conclusion

Helio is a real estate CRM designed to support high-stakes decision-making. By integrating search, analysis, and AI-driven insights into a single system, it reduces the complexity of finding and evaluating properties, particularly for users operating under time and information constraints.

Built With

Share this project:

Updates