I own a restoration company and use Xactimate daily for fire and water loss estimating.
I want to build an internal web-based AI tool that converts plain English scope notes into the closest matching Xactimate line item codes.
Example:
Input:
“Remove soot damaged tile floor in kitchen. 4 ft flood cut drywall. Replace 5 inch baseboard.”
Expected Output (structured JSON):
• FCT DEM
• DRY 4
• VNB R&R
This will be a private internal tool (not public SaaS at this stage).
Project Scope
The application should:
1. Accept paragraph-style scope notes
2. Break the text into individual scope tasks
3. Identify action types (DEM, REP, R&R, D&R)
4. Use semantic search to match each task to the closest line item from a provided Xactimate dataset
5. Return structured output (code, description, confidence score)
6. Flag ambiguous matches instead of hallucinating codes
The system MUST use a Retrieval-Augmented Generation (RAG) architecture.
The AI should NOT guess or invent codes. It must only select from the provided database.
⸻
Technical Requirements
Preferred Stack:
Backend:
• Python (FastAPI preferred)
• OpenAI API (GPT-4 class model)
• Vector database (Supabase, Pinecone, or similar)
Frontend:
• Simple web interface (React, Next.js, or similar)
• Text input field
• “Translate” button
• Results panel with structured output
Database:
I will provide a private export of:
• Line item code
• Description
• Category
• Region tag
This must be embedded and stored in a vector database for semantic retrieval.
⸻
Required Experience
You must have:
• Proven experience building RAG systems
• OpenAI API integration experience
• Experience with vector databases
• Experience designing AI guardrails to prevent hallucinations
• Ability to explain architecture clearly
Please include examples of similar AI systems you’ve built.
⸻
Deliverables
• Functional web-based MVP
• Vector search implementation
• AI prompt architecture
• Clear deployment documentation
• Source code ownership transferred to me
⸻
Timeline
3–6 weeks preferred.
⸻
Budget
$5,000–$10,000 depending on experience and proposal quality.
Open to fixed-price proposals.
Apply Now
Apply Now