The Challenge
Property assessors need valuations that are both accurate and equitable across different neighborhoods and property types. Traditional models produce static snapshots that miss market trends and can embed systematic bias — risking regulatory non-compliance and eroding public trust.
Approach
We built a living prediction system that accounts for temporal market shifts, detects and corrects for bias against IAAO compliance standards, and handles anomalous data that would otherwise skew results.
What We Built
State-of-the-art forecasting models capturing how property values shift over time
Bias detection and correction engineered directly into the model pipeline to meet IAAO standards
Robust anomaly detection to identify and manage unusual sales or data outliers
Four production pipelines: data ingestion, feature preparation, model training, and performance evaluation
Technologies & Methods
Selected Visuals
