Building a Real-Time Municipal Data Parser: How We Normalized 30 Years of Toronto Zoning By-Laws with AI

We’ll walk through how we built ZoningPal’s data ingestion and normalization engine that transforms scattered, inconsistent municipal zoning data into structured, query-able intelligence. This is a messy, real-world AI application dealing with:

PDF hell: Non-searchable scans, broken links, cross-references across 100+ documents
Data chaos: 30+ years of amendments, overlays, exceptions, and site-specific rules with zero standardization
Version drift: Multiple conflicting sources that don’t agree with each other
Accuracy requirements: Every answer needs to link back to the exact by-law clause

We’ll demo the live system, show actual running code, and share what worked (and what spectacularly failed) when building an AI system that municipal planners and architects trust.