Yehia Ahmed

- Trained multimodal probabilistic time series Transformer model for safety critical metrics and uncertainty on CARLA - ROS2 setup. - Enhanced safety assurance monitoring in autonomous driving systems through advanced DL and sequential modeling. - Data collection and parallel training using Slurm HPC cluster and use of containerization tools like docker.

- Conducted an in-depth review, reproduction, and evaluation of state-of-the-art literature on Out-of-Distribution (OOD). - Developed ML applications via Streamlit and pipelines focusing on Robustness, XAI, OOD, data imbalance, and other activities.

- Lead, designed, & implemented a new target AURIX TC3xx and safety mechanisms modules. - Built knowledge base, integrated, tested, & configured for a new MCAL layer while reporting bugs to suppliers. - Developed & configured, via Vector tools, project-specific features, RTE, OS, memory, and comm. stacks. - Implemented AUTOSAR modules using C/C++ compliant with Valeo & Automotive industry standards & practices. - Created automation scripts to improve code quality & save time across the company.

- Developed then integrated ML models into an ensemble one with Pytorch. Also, participated in Research activities.

- Designed, developed, tested, and documented secure open APIs on Django. Deployed and setup AWS servers.

GPA: 1.7 - Focus: Machine and Deep Learning, Computer Vision, Reinforcement Learning, and SW engineering - Authored a review paper on the state-of-the-art research in Input Dependent Randomized Smoothing Methods. - Courses: Computer Vision III, Machine Learning for Graphs & Sequential Data, Deep Learning , Machine Learning, Explainable AI for ML (xAI), Adv DL in Medicine (ADLM)

- GPA: 3.64/4.0 - Minor in Business Administration. - Courses: Practical Machine Deep Learning, Secure Sys., Distributed Sys., Embedded Sys., Software Eng, Object Oriented Programming (C++), Analysis & Design of Algorithms, Compiler Design. Extensive in C++. - Graduation Project: Lynks (Nodejs & Electron). Developed P2P decentralized distributed cloud storage system distributing data over many nodes across the network and using DHT, correction codes, and security encryption and design concepts.