Nancy Bou Kamel
Master Thesis (Data Science) @Cardo Systems, Ltd
Signup · Get unlimited contacts
WORK HISTORY
Master Thesis (Data Science) @Cardo Systems, Ltd
Straubing, DE
Master Thesis: A Lightweight Noise-Robust ASR-Commands Model for Battery-Powered Devices.Developed and deployed a highly efficient, noise-robust ASR command model for resource-constrained, battery-powered devices, significantly advancing performance and energy efficiency.1) Engineered and deployed a lightweight, noise-robust ASR command model (RNN Transducer) in Python, designed for battery-powered embedded devices, achieving 95% accuracy and 15% energy reduction.2) Enhanced model robustness and accuracy by 20% in low SNR environments by leveraging a custom RNN-T loss (Bypass RNN-T and Star RNN-T variants), Center Loss, and Weighted Finite-State Transducers (WFST) for improved phonetic distinction across 200+ distinct commands.3) Optimized model hyperparameters using Optuna, contributing to significant performance gains.4) Reduced model size to a compact 500KB via knowledge distillation, enabling seamless integration into resource-constrained systems with <5% performance loss.5) Surpassed a fine-tuned OpenAI Whisper model on proprietary data, demonstrating strong custom model development and achieving a 10% relative WER improvement.6) Managed transparent project progress through regular documentation and presentations, facilitating informed team collaboration.
EDUCATION
TU Dortmund University
Master's degree
Berlin University of Applied Sciences Berlin (BHT)
Master's degree
Lebanese International University
Bachelor's degree
ABOUT NANCY BOU KAMEL
As a Data Scientist and Cloud Developer with over 3 years of experience, I build and deploy innovative, end-to-end solutions. My expertise lies at the intersection of complex data science, scalable cloud infrastructure, and robust software development. I am skilled in transforming intricate data challenges into tangible business value through effective model deployment and optimization. My journey began in cloud engineering, where I architected and deployed secure, real-time data processing workflows on AWS, cutting deployment times by 50% and boosting data handling efficiency by 40%. This foundation in building reliable systems seamlessly transitioned into deep diving into data science. As a Machine Learning Engineer, I\'ve developed and deployed advanced deep learning models (Transformers, LSTMs) to forecast energy consumption with <2% MAPE, engineered noise-robust ASR command models achieving 95% accuracy on embedded devices, and significantly enhanced prediction accuracy through sophisticated optimization techniques. My work often involves large-scale data manipulation (10M+ rows), advanced NLP, and fine-tuning models like Whisper for superior performance. I\'m adept with Python and Java, PyTorch, TensorFlow, Docker, Kubernetes, and CI/CD, always striving to deliver solutions that are not just effective but also scalable, efficient, and robust. My passion lies in transforming intricate data into actionable insights and tangible products, always focusing on driving measurable business impact. I\'m eager to connect with opportunities where I can leverage my full spectrum of skills to solve challenging problems and contribute to cutting-edge advancements in AI, cloud, and software development.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.