Mohammed Geabel

A check-in system built to make screenshot-based attendance fraud impossible.

• Engineered HMAC-SHA256 signed QR tokens rotating every 90s with JWT sessions.
• Implemented GPS geo-fencing via the Haversine formula to validate proximity before granting attendance.

A full-stack C2C marketplace with escrow payments.

• Built a full-stack C2C platform with secure escrow payment, supporting 50+ listings across multiple transaction flows.
• Applied Clean Architecture and the Repository Pattern across 5+ service layers, reducing inter-module coupling.

A full-stack portal serving three data domains for non-technical staff.

• Built a full-stack portal for 3 data domains across a 5+ page React frontend with an admin CMS.
• Designed a RESTful API that reduced frontend data-fetch complexity for non-technical staff.

A wildfire hazard assessment platform combining satellite imagery and geospatial data analysis to track fire risk.

• Builds on TÜBİTAK-funded research applying deep learning to spatial weather data for wildfire risk prediction.
• Uses satellite imagery and geospatial intelligence to assess and monitor wildfire hazards.

A collaborative data mining course project: a shared preprocessing pipeline, then individual predictive models compared on a shared leaderboard.

• Collaborated on data collection, cleaning, and exploratory analysis to build a shared processed dataset.
• Independently designed and trained a predictive model, ranked on accuracy, explainability, and real-world applicability.

A data science project cleaning and exploring a 20,000+ listing Turkish housing dataset to generate pricing hypotheses.

• Cleaned and transformed a messy real-world dataset of 20,000+ Turkish homes, handling missing values and categorical variables.
• Ran exploratory analysis and feature engineering to generate pricing-related hypotheses and prepare the data for predictive modeling.