Hi, I'm Hussam

Tadwen: Arabic Handwritten Manuscript Recognition

An AI system that turns images of historical handwritten Arabic manuscripts into searchable, editable text. My part: collecting and preprocessing the training data.

• Python
• OpenCV
• Data preprocessing
• Computer vision
• Arabic handwriting recognition

Tadabur: Quran Recitation Speech Recognition

Fine-tuned OpenAI Whisper for Quran recitation: it transcribes the recitation, identifies the surah and ayah, and recognizes the reciter. Word error rate fell from 41.1% to 6.3%.

My role: Design decisions (implementation by Rakan Suliman)

• Speech recognition
• OpenAI Whisper
• Fine-tuning
• Arabic NLP

Customer Churn Prediction

Predicted which telecom customers will leave using four ML models in WEKA, then benchmarked them against a published big-data study. Best model: Logistic Regression, with 80.2% accuracy and 0.845 AUC.

• WEKA
• Classification
• Logistic Regression
• Random Forest
• Naive Bayes

Data Mining Coursework: Classification, Clustering and Association Rules

Three end-to-end WEKA assignments on seven UCI datasets: classification and regression, clustering, and association rule mining.

• WEKA
• Classification
• Clustering
• Association rules
• Apriori

Campus Energy Consumption Analysis

Analyzed 1.25 million hourly energy readings from 78 La Trobe University buildings to find usage patterns and support the university's 2029 net-zero goal.

• Data analysis
• Time series
• Statistics
• Random Forest
• Clustering

Tent Detector

Fine-tuned YOLOv11 to find tents in aerial and campsite imagery, reaching 98.3% [email protected].

SkillStack

A bilingual agentic study planner that builds your certification roadmap and puts it on your calendar.

Facial Emotion Classifier

A 4-layer CNN trained on FER2013 that reads 6 emotions live from a webcam with OpenCV.

Sumo Robotics Bot

Autonomous sumo firmware in C++ for Arduino with IR tracking and edge recovery. 2nd place.