Artificial Intelligence graduate · Makkah, Saudi Arabia

Ahmed Saeed
Al Yazid.

Artificial Intelligence · Machine Learning · Data Science · Computer Vision

I work on comparative computer-vision research and practical data systems. My portfolio brings together a team deep-learning study and a live Arabic educational platform.

Explore my work Download CV GitHub
3.37 / 4.00GPA
2026B.Sc. Artificial Intelligence
MakkahSaudi Arabia
Selected work

From an operational need to a live platform.

A deployed educational supervision platform puts forms, reporting, access control and Arabic workflows into daily use.

Live production project

Educational Supervision Platform

A production Arabic platform for educational supervision workflows in a private-school environment. It centralizes visits, records, reports, approvals and document exports.

01

Digitizes supervisory visit forms and structured teacher/supervisor records.

02

Provides dashboards, filters, visit metrics, reports, and operational workflows.

03

Exports official records to PDF, Word, and Excel while preserving Arabic RTL document requirements.

04

Runs on a cloud architecture using Cloudflare Pages & Workers, D1, Better Auth, and Backblaze B2.

Cloudflare PagesWorkersD1Better AuthBackblaze B2Arabic RTL
Open live platform Source repository is private to the school.
Graduation research

A controlled comparison, not just a headline accuracy.

This AI research-based team graduation project compared deep-learning architectures and transfer-learning strategies for glaucoma image classification under a controlled study protocol. It was a research comparison using synthetic images, not a clinical diagnostic product.

Team research project · 2025–2026

Comparative Deep-Learning Research for Glaucoma Classification

A comparative study of pretrained and from-scratch architectures with trustworthiness diagnostics.

The study compared six experimental conditions: ViT-Small and ResNet-18 under feature extraction and full fine-tuning, plus KAN-Transformer and a matched MLP-Transformer trained from scratch. All conditions shared a common evaluation protocol.

The dataset contained 4,828 synthetically generated retinal fundus images. Evaluation used stratified 3-fold cross-validation, discrimination metrics, bootstrap confidence intervals, calibration measures, peripheral-occlusion diagnostics, and Attention Rollout.

PythonPyTorchCross-validationCalibrationExplainable AI
96.3% Best accuracy

ResNet18-FT under the study protocol.

0.996 AUC-ROC

Reported for ResNet18-FT.

6 Conditions

Evaluated under a controlled comparative design.

4,828 Images

Synthetic fundus images across two classes.

1Compare architectures

Evaluate six conditions using stratified 3-fold cross-validation.

2Measure beyond accuracy

Report AUC-ROC, bootstrap confidence intervals and calibration alongside accuracy.

3Investigate model behavior

Use Attention Rollout and peripheral-occlusion diagnostics to inspect image reliance.

Academic projects

Additional machine-learning work.

Selected coursework in computer vision, regression, classification, preprocessing and data analysis.

CV

Animal Image Classification

Built a multi-class CNN image classifier with preprocessing and image augmentation.

TensorFlow / KerasCNNImage augmentation
ML

Diamond Price Prediction

Used diamond length, width and depth to build a linear regression workflow, evaluated with MAE, RMSE and R².

PandasScikit-learnRegression
DS

Obesity Risk Classification

Trained a Gaussian Naive Bayes model on selected features to classify obesity-risk levels and reviewed its classification report.

Scikit-learnGaussian Naive BayesClassification
Interactive lab

Put the diamond model to work.

Enter a diamond's three dimensions to run the linear regression from my coursework. I reproduced its coefficients from the original 500-row project dataset using the same 80/20 split and random seed as the notebook.

On the held-out test set: MAE $241, RMSE $329, R² 0.904. Estimates use dimensions only and are an experiment, not an appraisal or sale price.

01 / ORIGINAL PROJECT MODELDiamond price estimate
Use the sample values or enter your own dimensions.

The model runs in your browser. Dimensions are not sent or saved.

Technical toolkit

Tools I use to move from data to decisions.

My background combines AI experimentation, applied data work, and deployment of a real cloud-based platform.

Languages & Data

PythonSQLJavaPandasNumPyMatplotlib

Machine Learning

PyTorchTensorFlow / KerasScikit-learnCNNsVision TransformersResNetTransfer LearningXAI

Computer Vision & Evaluation

Image ClassificationCross-ValidationCalibrationAUC-ROCBootstrap CIs

Cloud & Delivery

Git / GitHubCloudflare PagesWorkersD1Better AuthBackblaze B2
Education

B.Sc. in Artificial Intelligence

Umm Al-Qura University · College of Computing

2022–2026GPA 3.37 / 4.00Makkah, Saudi Arabia
Certifications
  • Introduction to Artificial IntelligenceEdraak · 2022
  • Digital Skills in Artificial IntelligenceEdraak · 2022
Let’s connect

Looking for an AI, ML, or Data role?

I’m based in Makkah, Saudi Arabia. For Artificial Intelligence, machine learning, or data opportunities, contact me by email or phone.

ahmedalyzeed@gmail.com +966 55 848 9990 github.com/Ahmed-AlYazid