Construction Project Intelligence
AI-powered construction project management with document intelligence
Overview
Construction projects generate mountains of documents — RFIs, change orders, submittals, schedules. I built an AI assistant that understands construction documents, answers questions with citations, classifies RFIs automatically, and flags risks before they become problems.
The Problem
Construction projects generate vast amounts of documentation — RFIs, change orders, submittals, schedules, and budgets. Finding the right information at the right time is critical, but teams often rely on manual searching through unstructured document repositories, leading to delays and missed risks.
The Approach
Built a full-stack construction project intelligence platform with multi-project workspaces, role-based access control (6 role types), and an AI assistant using RAG-style document search with citations. The system includes RFI intelligence with NLP-based classification and entity extraction, plus analytics dashboards for risk indicators.
The backend uses FastAPI with PostgreSQL, featuring document upload with versioning, TF-IDF semantic search, and optional OpenAI embedding upgrades. The React frontend provides an intuitive interface for project managers, architects, and engineers.
Key Decisions
Designing for role-based access from day one ensured that sensitive project information is properly scoped. The vector store integration allows the AI assistant to retrieve relevant document snippets and answer questions with source citations.
Docker containerization ensures consistent deployment across development and production environments, with the architecture designed to be extensible for future BIM integration and real-time updates.
Impact
The platform provides a unified workspace for construction project management with AI-powered document intelligence, reducing the time spent searching for information and proactively identifying risks through RFI analysis and analytics dashboards.