AI Engineer · Beirut, Lebanon · Open to conversations
Building scalable data pipelines, semantic search, and production-ready LLM systems for organizations shaping policy across the Arab region.
About
A policy analyst has a question. The answer exists — scattered across thousands of documents and datasets, in different languages, in a repository nobody can search. Everything I’ve built lives in that gap: the pipelines that hold it all, and the machine learning that pulls the answer back out.
Education B.E. Civil & Environmental Engineering, Beirut Arab University (2015 — 2020). Honours student — CGPA 3.84 / 4.
Competitions First place, LIRA Competition — 3D concrete printing (2017). Our team printed mock-up concrete structures on a printer we built with mechanical engineering students, and formulated the mix that made it possible — high compressive strength without giving up workability, buildability, or setting time.
From multi-agent report generators to RAG advisors running inside government portals — I build LLM systems that reach production, not just demos.
Pipelines end to end: ingestion, validation, metadata, lineage. I build for the analysts who need the number to be right, and for the models that generate entire analyses on top of it.
My work powers policymaking tools across the Arab region — the portals economists search, the advisors ministries ask, the simulators policymakers run.
Featured Projects
Multi-agent LLM report generation
Developed a multi-agent LLM system that produces reports on demand, retrieving and synthesizing content from Manara’s document repository. Separate agents handle research, drafting, and quality control.
ESCWA’s first LLM policy advisor
Built ESCWA’s first LLM-powered policy advisor — a RAG assistant over Manara’s document and ADP’s dataset repository that answers in context, grounded in the source documents. It charts the data behind its answers, speaks through a realtime voice agent, and turns a topic into a generated podcast. Deployed as a standalone web app and integrated into the Manara and Arab Development Portal data catalog.
RAG analytics for Qatar’s Ministry of Labour
Built an LLM-powered RAG chatbot over government employment data, giving Qatar’s Ministry of Labour analytical answers that support evidence-based policymaking.
Data pipelines + Search Engine
Designed the architecture and built the pipeline ecosystem behind the data and search for the Manara and Arab Development Portal: pipelines for data ingestion, transformation, and loading, plus pipelines that embed dataset and document metadata and serve an optimized retrieval path through Azure AI Search. Also built LLM pipelines for metadata generation and document classification.
Optimization models for policymakers
Built the core engine behind the Index Simulator for Policymakers in the Arab Region, running the optimization models that simulate index and indicator values. Also contributed to the Financing for Development and Multidimensional Poverty Index models.
Agentic RAG over ESCWA’s technical cooperation archive
Built an pipelines that ingests ESCWA’s bilingual technical cooperation archive, indexes it for hybrid semantic search, and serves a citation-grounded chat assistant plus multi-document thematic summarization — both strictly evidence-bound to the retrieved documents.
District membership & dues portal
Built a full-stack Nuxt 3 and Firebase portal for Lions Clubs International’s Lebanon district, replacing spreadsheet-based membership tracking with a live database of the district’s club hierarchy and online dues collection through a hosted payment gateway.
Deep learning for macro forecasting
Led the data science team assessing whether a GDP forecasting model could reach 3% accuracy, testing deep neural networks against ensemble machine learning methods.
Experience
Mar 2023 — Present
United Nations ESCWA — ICT Section · Beirut, Lebanon
Aug 2021 — Mar 2023
United Nations ESCWA — ICT Section · Beirut, Lebanon
Nov 2020 — Aug 2021
United Nations ESCWA · Beirut, Lebanon
Mar 2024 — Aug 2024
Terascom · Beirut, Lebanon
Capabilities
Contact
I'm open to AI roles, and to collaborations on LLM products. Phone number is the fastest way to reach me.