Toronto Police Service - Global Search
(Note - This is not offically endorced by the Toronto Police Service)
Overview
Served as the Managing Architect to designe and implement a Global Search Platform for the Toronto Police Service (TPS) Business Intelligence & Analytics Unit, enabling officers to search across people, places, and things from a single unified interface. The engagement spanned the full solution lifecycle — from architecture and ingestion design through machine learning, security, and a production proof-of-concept — delivered through MC+A in partnership with Inmind Technologies and Attivio (aquired by ServiceNow).
The Challenge
TPS officers relied on multiple disconnected systems to locate information during investigations, requiring separate searches across internal records management systems, government databases, document repositories, and social media. This fragmented experience slowed investigations and left relevant evidence undiscovered.
Key challenges included:
- Unifying search across structured, semi-structured, and unstructured data sources — including legacy systems, CSV files, PDFs, audio, video, photos, SharePoint, and social media
- Connecting to both internal sources (Versadex, eCOPS, SuperG, file shares) and external government databases (CPIC, MTO)
- Meeting stringent security and internal affairs compliance requirements, including strict access control and audit trails
- Delivering machine learning-driven relevancy tuned for law enforcement use cases such as homicide investigation and business intelligence
Solution
Delivered a Proof of Concept (POC) implementation of the Attivio Cognitive Search Platform, demonstrating end-to-end global search across TPS data sources with custom ingestion, enrichment, and a demonstration UI and API.
1 - Federated Ingestion Architecture
Designed and built custom connectors to ingest and normalize data from Versadex (records management), eCOPS, SuperG, network file shares, and others. Defined ingestion workflows with enrichment stages for entity extraction, schema normalization, and profile building across all source types.
2 - Security & Access Control
Implemented a custom SAML authentication provider and Kerberos-based access controls to enforce row-level security aligned with TPS internal affairs requirements. Ensured that search results respected officer-level permissions across all connected data sources.
3 - Machine Learning & Relevancy
Developed use-case-specific relevancy models for Intel Operations, Business Intelligence, and Homicide investigation workflows. Applied query-time joins to surface related entities across documents and records, and built machine learning pipelines to bias ranking based on investigative context.
4 - Migration Path to Elasticsearch
After the initial Attivio implementation, designed a migration path toward Elasticsearch while preserving custom pipeline stages and SAML integrations, ensuring continuity as the platform evolved.
Technologies Used
- Attivio Cognitive Search Platform
- Elasticsearch
- NLP / Entity Extraction
- Custom SAML Provider
- Kerberos
- Query Time Joins
Results
- Delivered a successful POC adopted by investigators, improving cross-unit data access
- Officers gained a single search interface spanning 15+ years of records across internal and external systems
- Machine learning relevancy models reduced noise and surfaced contextually relevant results for active investigations
- Security architecture satisfied TPS internal affairs compliance requirements