
Farah Saadaoui
Ai Engineer
وظّف هذا الشخص
أرسل عرض عمل مباشرة لهذا المرشح
نبذة
AI Engineer specializing in LLM-based systems, agentic AI, and explainable computer vision. Proven experience developing production-grade AI systems and conducting research in medical imaging. Expertise in multi-agent systems, model robustness, and end-to-end ML pipelines, with research contributions under review.
الخبرة
AI Engineer at Vitafluence.ai (2025-05 – Present)
AI Engineer contributing to intelligent tutoring platform development and agent system enhancement
- Contributed to the development and deployment of TutorioGem, an intelligent tutoring platform, by architecting and integrating a Teaching Assistant Agent within a LangGraph-based system, enabling multi-step reasoning, improved interaction flows, and more accurate personalized learning support.
- Enhanced agent capabilities on the LakesAI platform by incorporating domain-specific scientific reasoning skills, refining prompt design, and improving context management, resulting in more accurate, coherent, and context-aware responses for complex queries.
- Improved agent orchestration by implementing fallback mechanisms (e.g., web search integration) and refactoring pipelines to reduce code duplication while maintaining system behavior, robustness, and performance.
- Participated in reviewing and improving system components by debugging and resolving issues in agent interactions and attachment handling, identifying root causes, and implementing fixes to enhance stability, reliability, and overall system performance.
Graduate Student Research Assistant at Intelligent Software Engineering (ISE) Lab - University of Michigan-Flint (2024-08 – 2025-04)
Conducted research on explainable CNN for skin lesion classification using dermatological datasets
- Developed and evaluated an explainable CNN for skin lesion classification using dermatological datasets such as ISIC, focusing on interpretable deep learning for medical decision support.
- Conducted extensive experiments on dermoscopic image classification using deep learning architectures and feature extraction pipelines, including preprocessing, augmentation, and dataset balancing to improve model robustness and representation quality.
- Evaluated model performance using clinically relevant metrics including AUC, accuracy, F1-score, sensitivity, specificity, and calibration error, ensuring comprehensive assessment of predictive reliability.
- Tested model robustness under dataset variability such as class imbalance, distribution shifts, and image quality variations, with emphasis on generalization to real-world clinical environments.
- Performed comparative analyses across multiple model configurations to assess stability, interpretability, and failure cases in medical image classification, supporting more reliable and explainable predictions.
- Work accomplished under the supervision of Prof. Marouane Kessentini (marouane@umich.edu) and Dr. Rania Khalsi.
- A manuscript based on this work is currently under review.
Undergraduate Research Assistant at Intelligent Software Engineering (ISE) Lab - University of Michigan-Flint (2024-01 – 2025-07)
Assisted in early development of skin lesion analysis project including dataset preparation and baseline experiments
- Contributed to the early development of a skin lesion analysis project by assisting in dataset preparation, including ISIC dataset curation, preprocessing, and quality filtering for dermoscopic images.
- Implemented initial baseline experiments for skin lesion classification using standard CNN architectures to establish performance benchmarks for later research extensions.
- Supported feature extraction experiments and exploratory analysis of dermoscopic image characteristics to assess data variability and class imbalance challenges.
- Assisted in designing the experimental pipeline used for subsequent explainable CNN development, including preprocessing workflows and training setup.
- Collaborated with the research team to document experiments and iterate on model design under the supervision of Prof. Marouane Kessentini and Dr. Rania Khalsi.
التعليم
Master of Science in Computer Science and Information Systems – University of Michigan-Flint (2024-08 – 2025-05)
Software Engineering Diploma in Software Engineering – Private Higher School of Engineering and Technology (2021-09 – 2024-07)
Preparatory Class in Engineering Studies – Preparatory Institute of Engineering Studies (2019-09 – 2021-06)


