cd ../projects
[ CASE_STUDY // 02 ]
Renovation Copilot
An AI assistant that turns a fixer-upper photo into a renovation plan with photorealistic previews and itemized costs.
OpenAIGoogle GeminiFastAPISSE StreamingPython
~30sRender Time
SSEStreaming
The Problem
Investors couldn't visualize what a fixer-upper could become. Hiring designers cost thousands of dollars and took weeks. They were guessing on renovation budgets, and bad guesses meant bad deals.
The Solution
I built an AI Copilot that turns a property photo and a few prompts into a complete renovation plan:
- Photorealistic previews generated via OpenAI and Google Gemini.
- Real-time streaming so the user sees output appear as it generates, instead of waiting for a long synchronous request.
- Line-item cost estimates broken down by room, material, and labor.
What used to be a multi-day, multi-person process became a 30-second AI conversation.
Stack
- Python / FastAPI orchestrator with strict Pydantic schemas at every API boundary.
- Server-Sent Events for streaming model output to the React frontend.
- OpenAI + Gemini as the primary model providers, with prompt routing logic to choose the right model per task.
- Cost-estimation engine in Python, fed by a curated material/labor catalog.
Why It Works
The product solves a real, measurable problem investors had. The AI part is just the implementation detail; the win is that someone can now stand on a property and price out a renovation in real time.