Applied ML & AI • Data Science • Policy
I build applied AI systems that turn messy data into decisions.
Applied AI for policy and decision support — turning disinformation and bias into clarity.
Interactive Terminal
Explore my work through a command-line interface. Type `help` to see commands. ↑/↓ for history, Tab for autocomplete.
help to get started.Work
Projects
Bias2 — UN Reports Hybrid Extractor & Bias Analyzer
FEnd-to-end system that turns UN SG reports on Lebanon into structured incidents, UNSCR legal mappings, and bias indicators for decision support.
- •PDF ingestion → Marker/Nougat → clean Markdown → paragraph JSONL.
- •Hybrid extraction → 50+ regex patterns + spaCy NER + targeted LLM calls.
- •Legal mapping → cached UNSCR (1701/1559/1325…) article text for fast cross-refs.
- •Bias analysis → Entman framing features + statistical reliability scoring.
- •Delivery → JSONL + Streamlit dashboard (search, filters, summaries).
Tracking and Countering Disinformation Campaigns in the MENA: A Data-Driven Approach
Analysis of 600k+ Arabic tweets on Egyptian politics to surface sentiment patterns, topic clusters, and signals of coordinated disinformation.
- •Kaggle dataset → merged raw JSON → Arabic text normalization & cleaning
- •Feature extraction → sentiment field, hashtags, user/time metadata
- •Topic modeling (LDA) → narrative clusters + coherence checks
- •Temporal patterns → spikes & co-occurrence for coordination signals
- •Visualization → timelines, topic bars, hashtag networks
Experience
Professional Background
JustAir — Data Science Capstone Intern
Sep 2025 – Present, Remote
- • Engineering anomaly detection & dashboards on municipal sensor data (PM2.5, PM10, O3, NO₂).
- • Integrating weather, traffic, and GIS layers for enriched predictions and policy-ready insights.
The Washington Institute for Near East Policy — Data Science & ML Intern
Jun – Aug 2025, Washington, D.C.
- • Built Bias2 pipeline for 114+ UN Secretary-General reports, mapping 2,300+ incidents into structured legal + bias indicators.
- • Developed hybrid extraction (50+ regex, spaCy NER, targeted LLMs) with UNSCR cache for fast cross-references.
- • Shipped Streamlit UI with reliability scoring and usage for policy briefings.
GPCR Modeling & Machine Learning Research — Research Assistant
Nov 2024 – May 2025, East Lansing, MI
- • Trained ML models with weighted ensemble sampling to uncover molecular features of ligand binding/unbinding.
- • Integrated docking pipelines (5-HT7, PDB 7XTC), feature engineering, and SHAP analysis for interpretability.
- • Automated protein–ligand workflows with Schrödinger Maestro & OpenEye, enabling scalable predictions.
Actuaris — Data Science Intern
Jun – Aug 2024, Casablanca, Morocco
- • Used ML models to predict Moroccan share index (MASI) trends with 15% higher accuracy.
- • Designed Python reporting pipeline processing 50k+ datapoints, cutting reporting time by 40%.
- • Presented portfolio optimization strategies to executives, including the CEO.
Renault — Marketing & Communication Intern
Mar – Apr 2019, Rabat, Morocco
- • Delivered marketing & communications support for Renault brand in Morocco, focusing on campaign outreach and digital presence.
About Me
Background & Focus
I'm an applied ML & AI engineer and a data scientist focused on decision-support systems. I like shipping pragmatic pipelines, clean interfaces, and measurable reliability. My work sits at the intersection of machine learning and policy analysis, where I build tools that help decision-makers navigate complex, data-rich environments with confidence.
Applied AI
Production-ready ML systems for real-world decision support.
Policy Analysis
Turning complex policy documents and frameworks into structured insights.
Data Engineering
Designing end-to-end pipelines that transform raw data into actionable intelligence.