triage
General-purpose risk modeling and prediction toolkit for policy and social good.
Build and evaluate machine learning pipelines for public policy and social good problems.
Triage is a Python-based machine learning toolkit designed to guide data scientists and analysts through complex design choices in predictive modeling. It suits teams working on early-warning systems, inspection prioritization, and social good problems who need a structured framework for generating features, training models, and conducting bias audits.
What it is
The toolkit provides an integrated interface to handle cohort selection, unit of analysis determination, outcome definition, and feature generation. It runs experiments via a command-line interface or a Python package, storing matrices and models locally or in S3.
Who it's for
It suits data scientists, analysts, and researchers building public policy prediction systems using PostgreSQL databases. It is not suitable for users looking for a no-code interface or those not working with relational event and geographical data.
Key features
Design Choice Guidance
Walks you through cohort selection, feature generation, and model training decisions.
Integrated Experiment Runner
Execute full machine learning pipelines via CLI or Python package imports.
PostgreSQL Integration
Pulls source data directly from a PostgreSQL 13+ database for training experiments.
Flexible Storage Options
Store generated matrices and models on your local disk or directly in Amazon S3.
Up and running in 3 steps
Our honest verdict
What's great
- Structures complex machine learning design choices into repeatable workflows
- Integrates database querying directly with experiment configuration files
- Supports running experiments through both a CLI and Python API
- Includes built-in database provisioning via Docker for testing
Where it falls short
- Requires a running PostgreSQL 13+ database loaded with source data
- Demands ample disk space to store generated matrices and model artifacts
- Setup involves installing specific C library dependencies like libpq-dev
Pricing & license
The only cost is running it yourself, as the toolkit requires your own PostgreSQL database and storage space for models.
Open remote AI-training roles
879 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
What database does Triage require?
Triage requires a PostgreSQL 13 or newer database containing your source events and geographical data.
Can I run Triage without installing PostgreSQL myself?
Yes, you can use the command 'triage db up' to create a vanilla PostgreSQL database using Docker.
How do I execute an experiment?
You can run experiments using the command line interface with your config file or import Triage as a Python package.
