Vahdat Abdelzad

Papers

Total Citations

H-Index

About

Vahdat Abdelzad is a researcher specializing in uncertainty estimation and safety-critical machine learning, with a particular focus on applications in autonomous driving and medical robotics. His work addresses one of the most pressing challenges in modern AI deployment: ensuring that predictive models can reliably communicate their own confidence levels, enabling safer decision-making in high-stakes environments. His most notable contribution, "Calibrating Uncertainties in Object Localization Task" (2018), tackles the problem of quantifying prediction uncertainty within object detection pipelines — a fundamental requirement for systems where errors can have life-threatening consequences. By developing methods to estimate the probability distributions of predicted objects, Abdelzad's research provides a critical bridge between raw neural network outputs and trustworthy, interpretable predictions suitable for real-world deployment. This work has garnered 19 citations, reflecting meaningful engagement from the robotics, computer vision, and autonomous systems communities. Abdelzad's contributions are particularly valuable as the field grapples with deploying deep learning models beyond controlled laboratory settings. Researchers and engineers working on perception systems for self-driving vehicles or robotic-assisted surgery will find his approach to uncertainty calibration both practically grounded and theoretically rigorous.