T-Risk: a disaster risk calculation engine based on Tensorflow
Year: 2026
Type: software-paper
Abstract: T-Risk is a fully differentiable, GPU-accelerated disaster risk calculation engine implemented in TensorFlow. Unlike traditional risk engines that rely on object-oriented programming (OOP) structures and Central Processing Unit (CPU) execution, T-Risk leverages tensor calculus and automatic differentiation to provide high-performance risk metrics and their gradients. This architecture enables the processing of large portfolios while offering novel capabilities in sensitivity analysis, model calibration, and portfolio optimization. The software is designed to run seamlessly on modern hardware accelerators (GPUs and TPUs), bridging the gap between traditional risk assessment and the emerging ecosystem of AI-driven high-performance computing. To facilitate usability, the engine works with a Streamlit-based web interface that handles data input, real-time validation, and interactive visualization of risk metrics and sensitivity gradients.
Source: SoftwareX
Author: Orlando Arroyo
Institution: Universidad Industrial de Santander
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $1,200
OpenAlex ID: W7161403903