AI-Powered Chatbot for University Services

Role: Machine Learning / NLP Engineer (7-person team) | Jan 2025 – Present Key Contributions: - Collected, cleaned, and structured a full Arabic training dataset covering 20+ student-service intents (courses, fees, admissions, transfers, scholarships, postgraduate, graduation, and more). - Built an end-to-end Arabic NLP pipeline: text normalization, Egyptian dialect to MSA mapping, intent classification, entity extraction (9+ entity types), follow-up question generation, and natural-language response generation. - Fine-tuned AraBERT on domain-specific queries and combined it with rule-based and TF-IDF classifiers in a hybrid architecture, achieving 92%+ accuracy on rule-based matches and 85%+ on the ML path. - Integrated the NLP service with a PHP/Laravel back-end and MySQL database through a FastAPI layer, supporting multi-tenant queries scoped by university and faculty. Problem Solved: 60% of student-affairs staff time was spent on repetitive queries, with 7–14 day application turnarounds and service limited to 9 AM–3 PM. Chat-MYU automates these interactions 24/7. Tech Stack: Python, FastAPI, AraBERT, Hugging Face Transformers, PyTorch, scikit-learn, NLTK, spaCy, Pandas, rapidfuzz, regex.