Build a Domestic Risk Assessment Tool for Triage: Data, Models, Privacy & Deployment | Sabina Firtala

How do you build a domestic risk assessment tool that meaningfully improves triage while protecting people’s privacy and avoiding bias? In this episode, Sabina Firtala from Frontline’s AI product development walks through the end-to-end process of building a domestic risk assessment tool for triage. Sabina brings hands-on experience across data wrangling, visualization, statistical testing, model training and validation, with a background in Natural Sciences and prior analyst roles in finance and SaaS, plus freelance work for mission-driven projects.

We cover problem framing and project scope, data sources (case management systems, public records, surveys), and data preparation: cleaning, linking and feature engineering. Sabina explains risk scoring and model architecture, evaluation metrics and bias assessment, and practical privacy, ethical and legal compliance measures. Deployment topics include integrating risk tools into frontline workflows, user interface and decision-support design, stakeholder training and trust, plus monitoring for model drift and alerts. The episode also addresses operational constraints, partnerships, funding and open resources. Listen for concrete guidance on building, evaluating and deploying a domestic risk assessment tool—focused on impact, fairness, privacy and sustainability.