About me
I’m passionate about AI, with several years of experience in developing cutting-edge AI solutions for leading companies such as Honda and Valeo. My expertise lies in Generative AI (LLMs, VLMs), robustness, and generalization. I earned my PhD under the supervision of Alexandre Alahi at EPFL and received my master’s and bachelor’s degrees in Electrical engineering from Sharif university of technology.
I’ve contributed to several AI projects that involved adapting large language models (LLMs) to NLP-specific tasks and Retrieval-Augmented Generation Systems (RAGs), as well as designing vision-based solutions using foundation models. During my Ph.D. at VITA, EPFL, my work focused on the robustness and explainability of machine learning models in the context of autonomous driving. I developed methodologies using adversarial attacks, synthetic data generation, and parametric perturbations to evaluate and interpret model performance. I also explored knowledge-based AI approaches and formal certification techniques to enhance model reliability. Additionally, during a six-month internship at Five AI (now acquired by Bosch) in Oxford, I worked on counterfactual-based explainable AI methods.
News
We have an exciting work accepted at CVPR 2025! The paper brings Certification, i.e., guaranteed robustness, to the trajectory prediction for the first time. Check it out here.
Our UniTraj paper got accepted to ECCV 2024! It is a comprehensive framework for trajectory prediction (over 200 stars so far). Check it out at UniTraj website.
Our social-attack paper got accepted to Transportation research part C (TRC) journal! Check it out s-attack library.’
Our scene-attack work got accepted to CVPR 2022! Check it out s-attack library.’
Do vehicle trajectory prediction models work “everywhere”? Check out our new work added to our s-attack library.