Job Description
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