I'm a Computer Science graduate from King Saud University, where I graduated with First Class Honors and a 4.9/5.0 GPA. My work spans machine learning, computer vision, and applied AI, with a focus on turning technical ideas into practical systems.
I've built models to detect undeveloped land across Riyadh using satellite imagery, developed an on-device voice emotion classifier for an iOS app I released on the App Store, and conducted AI research at ICTP in Italy exploring how neural networks can help explain decision-making in the brain. My work on satellite-based land detection also led to a peer-reviewed publication.
Across research and product development, I'm interested in what happens when AI moves beyond the model itself and becomes something people can use, learn from, or make decisions with.
Experience
Jun – Aug 2026
AI Researcher · ICTP
Selected as one of 10 students from Saudi Arabia for the TRIL Fellowship at the UNESCO-affiliated International Centre for Theoretical Physics in Trieste, Italy, through the KAUST Academy AI Specialization Program. Conducting research at the intersection of AI and neuroscience: extending RNN-based function approximation to handle continuous inputs, moving decision-making models beyond simplified discrete grids, with the goal of interpreting real neurological processes and explaining the actions behind them.
Recurrent Neural Networks (RNNs)
Neural Networks
Computational Neuroscience
Mathematical Modeling
Scientific Research
Naive Bayes
Jul – Aug 2025
AI Foundation Program · Apple Developer Academy
Selected as one of 66 participants in Saudi Arabia for the Apple Academy. Applied Challenge Based Learning to tackle workplace disengagement, a problem affecting an estimated 60% of employees and costing companies about $8.8 trillion a year. Designed the app in Figma and shipped Halah, an iOS mood and focus app built with Swift and Core ML for on-device inference. Presented the solution to Apple Academy mentors and international reviewers.
Dashboard homeAnalytics by districtMap of predicted vacant land
2025 – 2026 · King Saud University
TerraVision
Detects undeveloped land across Riyadh using satellite imagery and machine learning.
Inspired by Saudi Arabia's White Land fees policy, TerraVision uses satellite imagery and semantic segmentation to identify undeveloped land across Riyadh and track changes over time. I helped build the geospatial machine learning pipeline, from dataset preparation and annotation to augmentation and benchmarking 10 segmentation models. The resulting bilingual dashboard maps vacant land across Riyadh’s neighbourhoods and reports how much of each area is undeveloped, to support urban planning analysis.
Published
Peer-reviewed paper in Land (MDPI), 2026 · Scopus Q1
Model performance
95.6% IoU with SegFormer (MiT-B0), the best-performing model among 10 evaluated
Potential application
In discussions with Riyadh Municipality about adapting the project for city-wide land monitoring
HalahDiscover which planet you are todayCan’t tell? Halah reads the tone of your voiceFrom goal to done, easilyPick your focus styleEvery step brings you closerPeace of mind is part of productivityYour Halah journeyAn assistant that reviews your week
2025 · Apple AI Foundation Program, team project
Halah هالة
An iOS mood and productivity app that uses on-device AI to recognize emotion in your voice.
Built during the Apple AI Foundation Program, Halah turns daily mood check-ins into an interactive experience, representing detected emotions as different planets. I developed a voice-based emotion classification model trained on a dataset assembled from public speech datasets and deployed it directly on the iPhone using Create ML and Core ML. The app also includes speech-to-text, personalized AI-generated reflections, focus timers and routine tracking.
Shipped
Released on the App Store in 2025
Built in
Five weeks, from user research and design to a working iOS app
Selected
One of 66 participants in Saudi Arabia
SwiftUI
Core ML
Create ML
Machine Learning
Python
Speech Recognition
Summer 2021 · Tufts University
AaSpellchecker
Suggests the word you meant.
Built in Python as my final project for Tufts University's Introduction to Python Programming course. The spellchecker uses a vocabulary of approximately 66,000 words to identify misspellings and rank candidate corrections using edit distance and word frequency, similar to the suggestions behind a search engine's "Did you mean?" feature.
First Class Honors, GPA 4.9 / 5.0. Award for Academic Excellence (2023) for a 5.0 GPA and zero absences in the common first year; 3rd place in the BigOs competitive programming competition (2025).
Summer 2021
Tufts University, Massachusetts
Math 10: Introduction to Python Programming (MATH-0010-BA). Final project: the Spellchecker.
2019 – 2020
Saudi IOI National Team Trainee · Mawhiba
One of 20 trainees nationally for the International Olympiad in Informatics program; 200+ problems solved under contest time limits.
Certificates
Mathematics for Machine Learning and Data ScienceDeepLearning.AI, 60 hours
Python 3 Programming SpecializationUniversity of Michigan, 120 hours
Introduction to Data ScienceUniversity of Michigan, 34 hours
Forward ProgramMcKinsey & Company, 20 hours
Cybersecurity, Cryptology and ForensicsOxmedica, 112 hours
International Mathematical Olympiad trainingMawhiba, 180 hours
Contact
Open to AI engineering roles and research collaborations.
TerraVision
Halah