Master's in Data Science and AI from the University of Ottawa. I build an enterprise knowledge assistant on a local LLM and transformer-based anomaly detection over millions of network logs. These systems cut downtime, configuration errors, and time to root cause.
Systems I design, build, and run in production in Telecom Egypt's network, as AI & Data Engineer at Alkan Telecom. The code is internal, so each entry documents the architecture, the approach, and the results.
An enterprise assistant built on a local LLM, serving senior management and cross-domain teams. I designed the system architecture and resource plan, and implemented retrieval across internal documents, databases, and APIs using a Chroma vector store with hybrid retrieval (BM25 + embeddings + reciprocal rank fusion). The assistant replaces manual searches across scattered systems, and every answer stays on local infrastructure.
Multi-component ML project for network fault detection and root-cause analysis, processing millions of log messages ingested from Elasticsearch, with coverage scaled from 5 to 12 network devices. The system surfaces faults earlier and shortens the path from alarm to root cause.
Extracted and consolidated configuration templates from production routers into a standard golden-image configuration applied across devices, reducing configuration errors and outages and cutting setup time for new devices.
Designed and automated ETL and monitoring pipelines with Python, SQL, IBM CP4D, and DataStage, maintaining data integrity across large data volumes. Monitoring catches pipeline issues before they reach the teams that depend on the data.
Machine learning, NLP, and analytics projects, each with source available.
NLP system for detecting AI-generated text, delivered as a master's graduation project with Wakeb Data. Fine-tuned BERT and RoBERTa for English, AraBERT for Arabic, and mBERT for multilingual detection.
Sentiment analysis and drug-category classification over patient reviews: cleaning, tokenization, NER, and lemmatization, with EDA surfacing top conditions and drugs.
SVM and decision-tree models predicting patient no-shows at a medical center to improve appointment scheduling efficiency.
EDA over network traffic plus two classifiers, a tuned LDA baseline and a random forest, for robust DDoS attack detection.
Multi-tab Tableau dashboard analyzing customer demographics, product profitability, and sales trends for Adventure Works Cycles.
Interactive dashboard covering reseller performance, employee sales, and product trends, with drillthrough pages, date filters, and Base64 image handling for in-report visuals.
Automated ETL processes integrating student, course, department, and instructor data with SQL Server Integration Services.
Built and queried multidimensional cubes with MDX to extract sales insights from cubes and dimensions.
Time-series analysis of Seattle bicycle traffic identifying daily, weekly, monthly, and yearly patterns in Python.
Production AI engineering, team leadership, and teaching.
Used in production systems and analysis work.
Formal degrees alongside ongoing certification work.