Mohamed Ahmed

- Led the comprehensive development of a generative QA system, integrating Retrieval-Augmented Generation (RAG), Large Language Models (LLM), and Azure services to enhance first-level customer support. - Successfully projected a reduction in service team workload by at least 20% during the initial testing phase of the new system. - Successfully built a Task-Automation agent based on LLM fine-tuning that can perform simple tasks described in our software based on human-readable text.

- Contributed to the integration of a robot arm into a 3D printers cluster, achieving zero human-interaction and doubling printing requests using Python. - Integrated a 2D Computer-Vision Classification Model into the 3D printer cluster to ensure the completion and correctness of the printing requests using CNNs.

- Developed Auto-Encoders (AE) using PyTorch for dimensionality reduction on geometric shapes SVG dataset. - Implemented CNN-based Model to solve the nesting problem, optimizing spatial arrangements in manufacturing processes. - Enhanced the existing dataset by designing an automated data augmentation pipeline. - Successfully achieved an enhancement in the inference time of a Reinforcement-Learning agent by almost 80%, from 30 seconds to 6.5 seconds, per data point.

- Developed a web-based solution to streamline and automate essential paperwork processes for faster applicant approvals. - Optimization of Approval Processes: Collaborated with department managers to refine workflows, reducing process duration from one month to 4 days, well below the initial target of 7 days. - Technology Implementation: Leveraged a robust tech stack including Microsoft SQL Server, JavaScript, Node.js, and React.js to build and enhance the application’s functionality.

Dissertation: Leveraging Large Language Models for Human Intention Recognition and Next Action Prediction. Description: Fine-tuned LLaMa2, Mistral, and BERT models using HuggingFace, LoRa, and PEFT to recognize human intention in interactive systems. Relevant Modules: Theoretical and Methodological Foundations of Autonomous Systems, Machine Learning, Computer Vision, Machine Perception and Learning, Information Retrieval and Text Mining.

Dissertation Title: Drivable and Non-Drivable Area Detection Using 2D Camera for Autonomous Driving Description: Multi-classification for classes in semantic maps to identify the drivable area and non-drivable areas such as potholes and side-walks using CNNs and PyTorch. Relevant Modules: Mathematical Foundation(Calculus, Probability and Statistics, Discrete Math), Data Structures and Algorithms, OOP, Analysis and Design of Algorithms, Data Bases, ML, CV, Data Engineering, NLP.