Machine learning · data science · applied AI
I build fast, explainable AI products—from privacy-aware LLM evaluation to real-time sports analytics.
Building practical software for students. Independent products for academic planning, collaboration, and productivity.
Built independently by Gabriel Marquez.
From planning your next semester to finishing your next group project.
Academic planning · Flagship
Know where you stand. Plan what comes next.
A private academic planning workspace for UDST students. Turn a transcript into a clearer view of your GPA, degree progress, and the semesters ahead.
No account needed. Your academic records stay in your browser.
A data science foundation, a product-engineering instinct, and a bias toward evidence.
I turn messy questions into systems people can actually use—then measure whether those systems deserve their confidence.
My work moves between production ML, retrieval and privacy research, analytics, and full-stack product delivery. That range is deliberate: the best model is only useful when its data contract, evaluation, latency, and interface work together.
I graduated with honors in Data Science and Artificial Intelligence from the University of Doha for Science and Technology, with a cumulative GPA of 3.698 / 4.0. I am based in Doha and currently build sports analytics systems at Sports Vector.
Education
University of Doha for Science and Technology · With Honors · 2022–2026
Selected recognition
The technologies I use to move from experiments to reliable, usable products.
Training, evaluation, and explanation
Privacy-aware RAG and evaluation
Pipelines to stakeholder decisions
Full-stack delivery and production systems
Production engineering, applied research, and technical teaching across Doha and remote teams.
Sports Vector
Architecting production cricket analytics and RAG systems used by national teams and professional franchises.
University of Doha for Science and Technology
Supported programming, web development, and introductory data science labs across Python, JavaScript, HTML/CSS, SQL, and analytics tools.
UDST / Columbia-affiliated AI in Healthcare Program
Conducted literature review and framework analysis on AI adoption, readiness, governance, and healthcare-system integration across Qatar and MENA.
Cleano Laundry
Led a Next.js and Firebase operations dashboard used by 20+ vendors to manage orders, services, pricing, and performance analytics.
Production systems, privacy research, analytics, and full-stack AI products—each tied to a concrete outcome.
Privacy disparity benchmark for fine-tuned LLMs
A study-driven benchmark for membership-inference risk across multilingual causal language models, with calibrated low-FPR and subgroup disparity reporting.
Player valuation and shortlist system
A leakage-aware valuation pipeline with chronological splits, model comparison, SHAP explanations, fairness/error slicing, and Power BI-ready scouting exports.
Membership inference experiment package
A configurable mask-based privacy evaluation workflow for retrieval-augmented systems, packaged around YAML study definitions, resumable sweeps, structured artifacts, and automated reporting.
Multimodal indoor prediction API
Room and floor classification from 178 wireless access-point features with optional accelerometer, gyroscope, magnetometer, and heading inputs.
Attribute-controllable portrait generator
A compute-efficient InfoGAN and DCGAN hybrid trained on a custom footballer dataset, with DiffAugment, spectral normalization, mixed precision, and latent-space exploration.
Reproducible studies of privacy risk, evaluation quality, and responsible AI deployment.
Research focus
I build evaluation suites that surface risk across attacks, models, languages, and subgroups, with durable artifacts for independent review.
A unified empirical study of membership inference, attribute inference, data extraction, and backdoor attacks in language models and RAG systems.
Read on arXivResearch on AI adoption, implementation readiness, governance, interoperability, workforce development, and healthcare transformation in Qatar.
View profileNew releases, research milestones, and notes from my work in applied AI and sports technology.
Have a production ML problem, a privacy evaluation question, or an ambitious data product to build?
Send a short brief with the problem, constraints, and what success looks like. I’ll reply with a useful next step.
