Wissem Karous Portfolio

September 2024

Built an AI driven system to enable natural language querying and exploration of GitHub repositories. Leveraged Llama 3 LLM, applying fine-tuning on a dataset of code collected from GitHub to enhance code comprehension and generate accurate answers [GitHub].

Developed and deployed AI driven anomaly detection solutions for web security at Securas Technologies. Engineered advanced anomaly detection system using firewall journaling log data, achieving 99% detection rate for known attacks and 93% detection rate for unknown attacks through BERT-based models.

Started AI Engineer Internship at Securas Technologies, focusing on developing AI-driven anomaly detection solutions for web security using advanced BERT-based models and FastAPI deployment.

Developed an AI-driven system to analyze and secure automotive internal networks. Applied Active Automata Learning and LSTM models on time-series Electronic Control Unit (ECU) and CAN bus data to learn normal communication patterns and detect anomalies in real time [GitHub,Demo].

Worked on a smart city initiative to optimize traffic management and support infrastructure planning. Analyzed 20 months of traffic data from sensors at four major junctions, including irregular and sparse datasets. Built predictive models using LSTM networks and statistical forecasting [Kaggle].

Formulated a state of the art lipreading model with interactive web interface. Built and deployed a lipreading model (automatic speech recognition from visual cues on lip movements) with a Streamlit web interface, achieving lowest CER (0.6%) and WER (1.7%) since the first related publication in 2016 [Demo].
Crafted and implemented mobile solutions to improve operational efficiency at Business Software. Constructed a mobile app for attendance tracking using TensorFlow Lite, React Native, and OpenCV.
Served as Machine Learning Instructor & Department Lead at Google Developer Student Club ENETCOM, spearheading hands-on workshops and student projects in Machine Learning.
Delivered over 15 AI workshops as Mentor & Instructor at Microsoft Club Higher Institute of Applied Science and Technology of Sousse, covering Machine Learning, Deep Learning, Natural Language Processing, Large Language Models, and Recommender Systems.