Selected Work

The Lethbridge Police Service faced time-consuming, manual reporting processes that limited officer efficiency and delayed the flow of operational intelligence. There was a need for automated solutions that could streamline data interpretation and enhance reporting accuracy, while operating within a secure, network-isolated environment where data could not leave organizational control.

A secure, locally hosted GPU server was configured to run large language models entirely within the LPS network. A pilot program was designed to test four practical AI applications with select officers, gathering structured feedback before broader rollout.

Crime and Intelligence Analysts at LPS were manually searching multiple disconnected systems to gather situational intelligence, a process that was slow, inconsistent, and limited their ability to respond to emerging incidents with complete information. No unified search capability existed across the organization's heterogeneous data sources.

A Meta-Search platform was developed that indexes records from multiple source systems into a single, locally hosted search engine, providing a Google-like search experience across the full depth of LPS's organizational data.

Traditional hot spot analyses tend to surface the same persistent high-crime areas already known to experienced officers, providing little new operational intelligence. The actual need was for a capability to detect emerging, short-term crime spikes as they develop, enabling proactive resource deployment before a pattern becomes entrenched.

An anomaly detection algorithm was developed using kernel density estimation and z-score normalization, identifying crime concentrations that deviate significantly from historical baselines across 18 crime types and three years of historical data.

LPS officers needed immediate, consolidated access to actionable crime intelligence, but existing methods were fragmented across systems, required analyst intermediation, and were too slow to support real-time operational decisions. Intelligence that arrived after an officer left for patrol had no value.

A custom intelligence portal was built, integrating multiple analytical outputs, including predictive models, geospatial analytics, anomaly detection, and offender tracking, into a single, role-based interface accessible to officers on shift. Every module was designed around operational utility, not analytical completeness.

Human trafficking investigations require monitoring large volumes of online data across multiple platforms, a task that was manually intensive, inconsistent, and difficult to sustain at the required scale. Bylaw and enforcement officers needed a system that could automate monitoring and surface actionable intelligence without requiring constant manual triage.

Developed in collaboration with the City of Edmonton's Community Safety Supervisor, Project Exposure is an AI-powered intelligence dashboard that automates the collection and analysis of online indicators associated with trafficking activity.

IED attacks in Afghanistan posed a persistent threat to military operations. The analytical process used to link individual attacks to their probable source networks, by matching technical signatures, tactical patterns, and geospatial data, was rigorous but slow, taking up to three days per analysis. In an active operational environment, this delay had direct cost-of-life implications.

A nearest-neighbor hierarchical clustering algorithm was developed to analyze over 300 technical and tactical features from IED incident data, identifying attack series and linking them to probable source networks using TF-IDF style feature weighting to prioritize the most discriminating characteristics.

Traditional crime analysis focuses on what happened and where, but rarely on why. Without understanding the social conditions, environmental influences, and historical patterns that drive criminal activity, prevention strategies tend to be reactive rather than structural. The City of Edmonton needed a more contextual approach to crime analysis to guide targeted community investment.

The Contextual Analysis of Crime platform was developed in 2015, integrating social, environmental, and historical data alongside traditional crime data. Advanced analytical techniques were applied to surface the root factors driving crime trends, enabling prevention strategies that targeted underlying causes rather than surface indicators.

Crime patterns in urban transit systems are dynamic and location-specific, making static resource allocation inefficient. Traditional hot spot methods were insufficiently accurate and too slow to be operationally useful; officers were frequently deployed to areas based on historical averages that no longer reflected current risk patterns.

A fully automated predictive analytics tool was developed that analyzed historical crime data to generate daily, hourly crime forecasts at specific transit locations, regenerated automatically every day from current data, with no analyst intervention required between cycles.