ADYou

Inspiration

California's severe housing crisis needs solutions, and millions of backyards are sitting empty. But building an Accessory Dwelling Unit (ADU) is a permitting, zoning, and financial nightmare. We built ADYou: an AI assistant for your backyard that democratizes real estate development for everyday homeowners.

What it does

ADYou is an end-to-end feasibility, finance, and marketplace platform.

Instant Feasibility: Search an address, fetch exact house boundaries, and draw your ADU on a 3D map to instantly calculate buildable zones and state setbacks.
Context-Aware AI Advisor: Chat with a RAG-powered bot trained on the CA HCD Handbook that already knows the exact lot size and footprint you just drew. It asks personalized questions that it uses to generate requirements and tips based strictly on the PDF.
Financial Predictor: Get instant build cost ranges, projected rent, and 10-year break-even timelines based on your city, build type, and current market.
ADU Marketplace: A dedicated, two-sided rental platform. Prospective tenants can seamlessly browse, filter, and save ADU listings, while homeowners use our AI Canvas Editor to auto-generate optimized property listings.
Report Generation: Curates a comprehensive report based on chat sessions. This contains a financial breakdown, description of the ADU, and recommendations + next steps.

How we built it

Frontend: Next.js (App Router), React, and Tailwind. We utilized on Mapbox GL, Mapbox Draw, and Turf.js for complex, browser-side spatial math.
Backend: A lightning-fast FastAPI (Python) engine utilizing Shapely and pyproj to handle complex geospatial coordinate projections.
AI/RAG Pipeline: We built a custom local vector index using sentence-transformers, powered by Groq (Llama 3) and Google Gemini for ultra-low latency inference.

Challenges we ran into

Spatial Math: Translating spherical web-mercator coordinates (lat/lng) into flat state-plane geometry to accurately calculate 4-foot legal setbacks.
Context Handoffs: Seamlessly passing complex 3D polygon data and user-onboarding context from the Next.js map directly into the RAG system prompt without losing state. Keeping our RAG agent persistent to keep all context from the chat.
Zero Hallucinations: Tuning our RAG retrieval to guarantee the LLM only cites the CA ADU Handbook, avoiding illegal or hallucinated zoning advice.

Accomplishments that we're proud of

Solving a Real-World Crisis: We identified a massive, costly bottleneck in California's housing market and built a practical, scalable solution that empowers everyday homeowners to become housing providers.
Component Integration: We divided and conquered across distinct, complex domains - frontend UI, geospatial mapping, backend RAG pipelines, and marketplace logic - and successfully merged our separate components into a single, cohesive architecture in just 48 hours.

What we learned

The Hidden Complexity of ADUs: Building an ADU is deceptively complex.
AI Agent Orchestration: We learned how to design and architect intelligent LLM routing agents. We mapped out how a master agent can classify user intent in real-time.
Advanced Geospatial Web Tech: Learning browser-side mapping and geometric calculations using Mapbox Draw and Turf.js to create a highly responsive user experience.
Lean Backend Engineering: Building local vector databases from scratch without expensive enterprise tools, and managing modern Python backends using uv for blistering fast API development.

What's next for ADYou

Total Regulatory Automation: Scaling our compliance engine to automatically ingest and process hyper-local municipal codes across all of the United States (outside CA).