Mohamed Hatem AI Team Lead
العربية , Read this page in Arabic

Mohamed Hatem, AI Team Lead at Globant. I build multi-agent LLM systems that hold up in production, in Arabic and English.

AI Team Lead at Globant with 7 years in data and AI. I built a 14-agent assistant for the stock-media platform ArabsStock and, at Wakeb Data, a 6-agent bilingual chatbot whose self-hosted models I ran on A100 GPUs.

See my work Download CV

Open to new roles · remote, in Cairo, or relocating to the Gulf or Europe in Cairo

An illustration of the pattern behind the agent systems I build. guard screens the input; the orchestrator routes the request to retrieve for hybrid search and to rerank to keep the best context; generate writes the answer; evaluate checks its quality. When live data is needed, the orchestrator calls tools over MCP. Not every system uses every step.
Use the arrow keys to move between agents. Press Enter for details.

The pattern behind the agent systems I build. Select a node.

years in data and AI, now leading an AI team
7
agents in one LangGraph assistant
14
agents in a bilingual production chatbot
6
of stock assets made searchable
Millions
GPUs serving Qwen through vLLM
A100

Shipped at

01 What I build

Production AI, from the agent graph down to the GPU.

I own the whole path: how agents route, what they retrieve, where the model runs, and how we know the answers are right.

Multi-agent LLM systems

LangGraph orchestration with command-based routing and modular subgraphs: from a 6-agent bilingual chatbot to a 14-agent assistant for a stock-media platform.

LangGraph · MCP · OpenAI API · Qwen · async Python

Retrieval and search

Hybrid vector + BM25 retrieval with LLM reranking, and semantic and visual search over millions of media assets.

Pinecone · OpenAI embeddings · BM25 · LLM reranking

LLM serving and MLOps

Qwen on A100 GPUs with vLLM: model versions, GPU allocation, health checks and automated restarts. Dockerised FastAPI and gRPC services, CI/CD.

vLLM · Docker · Nginx · gRPC · CI/CD

Evaluation and observability

LLM evaluation, tracing and monitoring, so we know a model is right before users find out it isn't.

staged-llm-evaluator · prompt-reliability-harness · observability

Arabic and English

Most of what I have shipped is bilingual: Arabic and English in the same product, from retrieval and search to the interface.

Arabic NLP · RTL interfaces · bilingual search

Front end, too

Accessible, fast, multilingual sites with Astro, TypeScript and Tailwind CSS, including Arabic RTL. See the logic-controls.com case study.

Astro · TypeScript · Tailwind CSS · Cloudflare Workers

02 Selected work

Systems I've built and run

Work since 2020, from multi-agent assistants to data pipelines. Client details stay at the level each employer allows.

Globant AI Team Lead 2026

Saudi Tourism Authority: Noura for visitors and an AI assistant for staff

Led the AI engineering for two Saudi Tourism Authority assistants on one LangGraph agent: Noura for visitors and a staff assistant, in Arabic and English.

  • One agent serves both Noura and the staff assistant, kept apart by the audience gate.
  • Noura was accepted in the client's user-acceptance testing (UAT) in September 2026.
  • Every action waits for the user's approval, and answers show their sources.
  • LangGraph
  • Python
  • FastAPI
  • RAG
  • LLM evaluation
Wakeb Data during 2024–2026

A 6-agent bilingual chatbot on self-hosted Qwen

Architected a fully async 6-agent LangGraph chatbot in Arabic and English, and ran its Qwen models on vLLM across production A100 servers.

  • LangGraph
  • Qwen
  • vLLM
  • NVIDIA A100
Freelance Logic Controls 2026

logic-controls.com: a 4-language site, accessible and tested end to end

Freelance build of logic-controls.com in English, Arabic (RTL), Chinese and Turkish: Astro on Cloudflare Workers, WCAG 2.2 AA, Lighthouse 98–100 mobile.

  • Live at logic-controls.com in 4 languages.
  • WCAG 2.2 AA with zero axe violations; Lighthouse 98–100 on mobile and 100 on desktop.
  • About 950 unit and 900 end-to-end tests run in CI.
  • Astro
  • TypeScript
  • Tailwind CSS
  • Cloudflare Workers
  • Turnstile

More work

Wakeb Data during 2024–2026

AIP semantic search, version 2

Designed the second version of the AI semantic search for the flagship AIP project: more accurate retrieval, faster search, better image search.

Tabaqat 2023–2024

GIS data pipelines and map serving

Designed and ran all data pipelines in Prefect (ETL for Saudi government GIS data), queried Overture Maps with Athena and DuckDB, and deployed map serving.

03 Career

Experience

7 years, from SQL at a payments company to leading AI engineering at Globant.

Role

AI Team Lead Globant  · Apr 2026 – Present

type
Full-time, Remote
where
Remote, from Cairo
Tech
LangGraph, Agentic LLM runtime, Retrieval, Human-approval gates, Evaluation
  • Leading AI engineering on the Saudi Tourism Authority (STA) project: an agentic LLM runtime (LangGraph) with retrieval, human-approval gates and evaluation, serving Noura, the public visitor assistant, and an internal staff assistant, in Arabic and English.

Senior AI & MLOps Engineer Wakeb Data  · Sep 2024 – Apr 2026

type
Full-time, Hybrid
where
Giza, Egypt
Tech
LangGraph, Qwen, vLLM, NVIDIA A100, RAG, BM25, LLM reranking, MCP, FastAPI, gRPC, Docker, Nginx, SQLite, CI/CD
  • Architected and deployed a 6-agent bilingual (Arabic/English) chatbot in LangGraph: fully async, built from modular subgraphs, with agents routed by LangGraph's Command instead of fixed edges.
  • Owned the model-serving lifecycle for Qwen on vLLM: serving configuration, A100 GPU allocation, model versions, health monitoring and automated restarts across multiple production GPU servers.
  • Built hybrid retrieval for RAG (vector + BM25, reranked by an LLM) with SQLite node-level caching and web search through the Brave Search MCP server, shipped as Dockerised FastAPI services behind Nginx.
  • Designed version 2 of the AI semantic search for the flagship AIP project: more accurate retrieval, faster search and better image search.
  • Built a bilingual enterprise assistant (Arabic/English) for a Saudi government client, with LLM evaluation and observability to monitor answer quality.
  • Developed the Wakeb Papers Research Agent for searching and analysing academic papers, with agent memory and multi-turn conversational context.
  • Migrated key APIs from REST to gRPC to cut latency between microservices.
  • Ran CI/CD and containerised deployments (Docker, Git, shell-scripted automation) across development and production.

AI Engineer ArabsStock  · Apr 2023 – Dec 2024

type
Full-time, Remote
where
Remote (Dubai, UAE)
Tech
LangGraph, OpenAI API, Pinecone, Python, FastAPI, PostgreSQL, Docker, vLLM, NVIDIA A100
  • Architected and built the ArabsStock Assistant: 14 specialised agents orchestrated with LangGraph for semantic search, AI image generation, database operations, content moderation and other AI services across the platform.
  • Built the semantic search pipeline (OpenAI embeddings + Pinecone) with automated batch ingestion, indexing millions of royalty-free photos, videos and vector graphics.
  • Added visual-similarity image search across the whole media library.
  • Took the platform into audio: sound generation and search for a beta sound library, instrument classification, BPM detection and audio watermarking. The model endpoints ran as containerised FastAPI microservices.
  • Set up LLM agent evaluation to track reliability, response accuracy and agent routing in production.
  • Built production pipelines for raw video and audio, including footage from ArabsStock studios: quality optimisation and metadata enrichment.

Software Engineer Tabaqat  · Feb 2023 – May 2024 , Part-time

type
Part-time, Remote
where
Remote (Saudi Arabia)
Tech
Prefect, AWS Athena, DuckDB, Overture Maps, MapProxy, Kepler.gl, MapLibre, MBTileServer
  • Designed and managed all data pipelines in Prefect, including ETL for GIS data from Saudi government sources.
  • Queried geospatial data from Overture Maps with AWS Athena and DuckDB for spatial analytics.
  • Deployed map serving and visualisation with MapProxy, Kepler.gl, MapLibre and MBTileServer.

Data Scientist Empoweromics  · Jan 2022 – Mar 2023

type
Full-time
where
Cairo, Egypt
Tech
Python, PyTorch, NLP, Computer vision, Azure Functions, Cosmos DB, SQL Server, Firebase, Flutter
Award
Best Engineer of the Year
  • Designed a data integration system that merged data from 3,008+ real-estate developers, each with its own schema, to scale the company's real-time e-map.
  • Built classification models for 7 client lifecycle stages (Inactive, Assigned, In Progress, Lost, Won, Void, Active) so brokers could prioritise high-value leads.
  • Built automatic WhatsApp message classification for text and images (NLP + computer vision), routing each message to Tech, HR, Accounting, Operations or Marketing.
  • Built Smart Search: fuzzy client lookup that finds the right record from misspelled or partial Arabic and English names.
  • Trained a deep-learning model in PyTorch that infers gender from Arabic names, to sharpen client targeting and campaign personalisation.
  • Designed ETL from Cosmos DB (NoSQL) to Microsoft SQL Server, scheduled with Azure Functions.

AI Research Assistant The British University in Dubai · Umm Al-Qura University  · Mar – Nov 2022 , Part-time

type
Part-time, Remote
where
Remote (Dubai, UAE)
Tech
Machine learning, Deep learning, NLP, Recommender systems
  • Primary research assistant to two master's students in biomedical informatics at The British University in Dubai, an engineer and a pharmacist.
  • Gave lectures and training sessions on linear algebra, probability and statistics, machine learning, deep learning, data mining, recommender systems and natural language understanding.
  • Mentored master's research at BUID and Umm Al-Qura University on text summarisation, recommender systems and classification models.

AI Instructor Military Technical College  · Aug 2022 , Part-time

type
Part-time
where
Cairo, Egypt
Tech
Machine learning, Deep learning
  • Delivered specialised AI courses to engineers in the Egyptian Armed Forces: machine-learning fundamentals, deep learning and practical deployment.

Freelance Data Scientist AestheticsPro (US healthcare software)  · Feb – Apr 2022 , Freelance

type
Freelance, Remote
where
Remote (Washington, US)
Tech
Web scraping, SQL Server, Exploratory data analysis, Dashboards, Machine learning
  • Scraped and cleaned healthcare data and loaded it into SQL Server for structured storage and retrieval.
  • Explored the data and built a detailed dashboard for stakeholders.
  • Built ML models that flag patients likely to leave, used to target retention offers.

SQL Developer Misria Digital Payments (MDP)  · Jan 2020 – Jan 2021

type
Full-time
where
Cairo, Egypt
Tech
SQL, Stored procedures, Query tuning, RDBMS
  • Wrote and optimised complex SQL stored procedures for production payment systems.
  • Tuned large SQL statements, stored procedures and views to cut execution time, and looked after database integrity and security.

M.Sc. in Data Science Cairo University  · Sep 2023 – 2027 (expected) , Study

where
Cairo University, Faculty of Graduate Studies for Statistical Research
notes
In progress, expected 2027
courses
Advanced Machine Learning, Advanced Bayesian Inference, Pattern Recognition, Deep Neural Networks, Time Series Forecasting, Data Science Decision Support Systems, Recommender Systems (project)

Diploma in Artificial Intelligence Information Technology Institute (ITI)  · Apr 2021 – Feb 2022 , Study

where
Information Technology Institute (ITI)
notes
9-month programme in partnership with EPITA, Paris
Use ↑ ↓ to move between roles, Enter to open.
04 About

About me

Portrait of Mohamed Hatem
Based in
Cairo, Egypt
Time zone
UTC+2 (UTC+3 in summer) in Cairo
Languages
Arabic (Native) · English (Professional working proficiency)

I'm an AI Team Lead at Globant, based in Cairo. I design, ship and run production LLM systems: multi-agent orchestration, retrieval, self-hosted model serving, and the evaluation and monitoring that keep them honest. At Globant I lead the AI engineering for the Saudi Tourism Authority's assistants: Noura for visitors and an assistant for staff.

I came to AI through data. I started in 2020 writing SQL for a payments company, moved into data science, then built AI platforms in production. Before Globant I architected a 14-agent assistant for ArabsStock and, at Wakeb Data, a 6-agent bilingual chatbot whose Qwen models I served with vLLM on A100 GPUs.

Most of what I ship works in Arabic and English. I work remotely with teams in the Gulf and the US. I earned the AWS Certified Machine Learning – Specialty in 2022 and I'm completing an M.Sc. in Data Science at Cairo University (expected 2027).

Open to
AI/ML Team Lead, Senior AI Engineer, MLOps Engineer, AI Agents / LLM Engineer and ML Engineer
Work arrangements
  • Remote, from Cairo
  • Relocation to the Gulf Saudi Arabia, UAE, Qatar and nearby
  • Relocation to Europe
  • Hybrid or on-site in Cairo
05 Code

Open source and client work

Small, focused tools for LLM work: grading outputs, testing prompt reliability, serving retrieval. Each README says what the code does and what it doesn't.

  • A replayable LangGraph pipeline that grades LLM outputs in separate quality, safety and retry stages, then computes the pass / review / fail verdict in plain Python, not with the model.

    PythonLangGraphOpenAI APIpytest MIT licence

  • Measures how consistently an LLM answers when the same question is paraphrased, misspelled, padded with noise, reordered or given distractors.

    PythonOpenAI APIEmbeddings MIT licence

  • A small retrieval-only FastAPI service: OpenAI embeddings, cosine and dot-product indexes side by side, a guardrail in front, and latency and hit-rate metrics.

    PythonFastAPIOpenAI embeddingspydantic-settings MIT licence

  • Arabic sentiment classification on about 455k tweets and reviews, comparing recurrent models (BiGRU, BiLSTM), CNNs on AraVec embeddings and classical ML.

  • A Dash app that summarises a news article or URL with DistilBART, T5 and TextRank side by side, showing time and length for each.

    Team project, ITI AI diploma (2022)

  • Source of logic-controls.com: a 4-language Astro site on Cloudflare Workers with an interactive HMI demo.

Where I started

Coursework from my AI diploma (2021–22), kept public as a record.

All repositories on GitHub
06 AI assistant

Ask about my work

My assistant has read my CV, LinkedIn and GitHub. Ask it about my experience, or paste a job description to check the fit. It answers in English or Arabic.

  1. Your question chat widget, EN or AR
  2. Worker Cloudflare, /api/chat, rate-limited
  3. Gemini whole profile in context, no vector DB
  4. Answer streamed back as it is written

How it works: one knowledge file, a Cloudflare Worker and Google Gemini. No vector database; the whole profile fits in the model's context.

How it's built
07 Toolbox

What I work with

Grouped by the job each tool does. Endorsement counts are from LinkedIn.

Skills, with receipts

Every dot marks where I used the skill for real: in a role, in freelance work or in a public repository.

Skills by where they were used. A filled dot means the skill was used in that role or project.
Skill Globant Wakeb ArabsStock Tabaqat Empoweromics Freelance & earlier Open source
Agents and LLMs
Multi-agent orchestration no yes yes no no no no
LangGraph yes yes yes no no no yes
MCP tool integration no yes no no no no no
OpenAI API no no yes no no no yes
Retrieval
RAG pipelines yes yes no no no no yes
Hybrid BM25 + dense, LLM reranking no yes no no no no no
Semantic and visual search no yes yes no no no no
Serving and MLOps
LLM serving (vLLM, GPUs) no yes yes no no no no
FastAPI microservices no yes yes no no no yes
Docker, CI/CD no yes yes no no yes no
Workflow orchestration (Prefect) no no no yes no no no
Quality
LLM evaluation and monitoring yes yes yes no no no yes
Accessibility and testing (front end) no no no no no yes yes
Data and ML
NLP (Arabic and English) no yes no no yes no yes
Computer vision and audio ML no no yes no yes no no
ETL and SQL no no no yes yes yes no
Cloud (AWS, Azure, Cloudflare) no no no yes yes yes no

Agents and LLMs

  • Multi-agent orchestration , Used in: Wakeb , ArabsStock
  • LangGraph , Used in: Globant , Wakeb , ArabsStock , Open source
  • MCP tool integration , Used in: Wakeb
  • OpenAI API , Used in: ArabsStock , Open source

Retrieval

  • RAG pipelines , Used in: Globant , Wakeb , Open source
  • Hybrid BM25 + dense, LLM reranking , Used in: Wakeb
  • Semantic and visual search , Used in: Wakeb , ArabsStock

Serving and MLOps

  • LLM serving (vLLM, GPUs) , Used in: Wakeb , ArabsStock
  • FastAPI microservices , Used in: Wakeb , ArabsStock , Open source
  • Docker, CI/CD , Used in: Wakeb , ArabsStock , Freelance & earlier
  • Workflow orchestration (Prefect) , Used in: Tabaqat

Quality

  • LLM evaluation and monitoring , Used in: Globant , Wakeb , ArabsStock , Open source
  • Accessibility and testing (front end) , Used in: Freelance & earlier , Open source

Data and ML

  • NLP (Arabic and English) , Used in: Wakeb , Empoweromics , Open source
  • Computer vision and audio ML , Used in: ArabsStock , Empoweromics
  • ETL and SQL , Used in: Tabaqat , Empoweromics , Freelance & earlier
  • Cloud (AWS, Azure, Cloudflare) , Used in: Tabaqat , Empoweromics , Freelance & earlier

Freelance & earlier: Logic Controls (front end), AestheticsPro, Misria Digital Payments.

The full toolbox

  • LLMs & agents

    Multi-agent orchestration in production, from 6 to 14 agents.

    • LangGraph
    • Multi-agent systems
    • OpenAI API
    • Qwen
    • MCP (Model Context Protocol)
    • Agent memory
    • Prompt engineering

    Also worked with: LangChain, Azure OpenAI, Whisper, BLIP

  • Retrieval & search

    RAG and semantic search, from hybrid BM25 + dense retrieval to visual search over millions of assets.

    • RAG pipelines
    • Hybrid search (vector + BM25)
    • LLM reranking
    • Semantic & visual search
    • Pinecone

    Also worked with: ChromaDB, Qdrant, pgvector

  • LLM serving & MLOps

    Self-hosted models on GPUs, kept healthy in production.

    • vLLM
    • NVIDIA A100 GPU serving
    • Model versioning & registry
    • Health monitoring & auto-restart
    • Docker & Docker Compose
    • CI/CD
    • Nginx
    • Prefect
    • Linux & shell scripting

    Also worked with: Kubernetes (AWS EKS), Celery

  • Evaluation & observability

    Knowing a model is right before users find out it isn't.

    • LLM evaluation
    • Observability & tracing
    • Prompt robustness testing
    • Guardrails

    Also worked with: Langfuse

  • Backend & data

    Python services and the databases behind them.

    • Python (10 endorsements on LinkedIn)
    • FastAPI
    • gRPC
    • PostgreSQL (11 endorsements on LinkedIn)
    • SQL (9 endorsements on LinkedIn)
    • SQL Server
    • Cosmos DB
    • DuckDB
    • ETL pipeline design

    Also worked with: MongoDB(12 endorsements on LinkedIn), Go

  • Machine learning & data science

    The foundation: models, features and the analysis around them.

    • Artificial intelligence (19 endorsements on LinkedIn)
    • Machine learning (17 endorsements on LinkedIn)
    • Deep learning (17 endorsements on LinkedIn)
    • Data science (15 endorsements on LinkedIn)
    • Data analysis (15 endorsements on LinkedIn)
    • Data visualisation (12 endorsements on LinkedIn)
    • PyTorch
    • TensorFlow
    • scikit-learn
    • NLP (Arabic and English)
    • Computer vision
    • Spark

    Also worked with: Algorithms(16 endorsements on LinkedIn)

  • Cloud

    AWS Certified ML – Specialty (2022); Athena on AWS at Tabaqat; Functions and Cosmos DB on Azure at Empoweromics.

    • AWS (SageMaker, EC2, S3, Athena, EKS)
    • Azure (Functions, Cosmos DB, DevOps)
    • Cloudflare Workers

    Also worked with: GCP

  • Frontend

    Accessible, multilingual sites with Arabic RTL, tested end to end.

    • TypeScript
    • Astro
    • Tailwind CSS
    • Accessibility (WCAG 2.2 AA)
    • Arabic RTL & internationalisation
    • Testing (Vitest, Playwright)

  • Ways of working

    Leading an AI engineering team; earlier, teaching and mentoring at university level.

    • Team leadership
    • Teaching & mentoring

    Also worked with: Scrum(15 endorsements on LinkedIn), Agile(15 endorsements on LinkedIn), System design

08 Learning

Education & certifications

Education

  • M.Sc. in Data Science
    Cairo University, Faculty of Graduate Studies for Statistical Research

    In progress, expected 2027

    Coursework: Advanced Machine Learning, Advanced Bayesian Inference, Pattern Recognition, Deep Neural Networks, Time Series Forecasting, Data Science Decision Support Systems and Recommender Systems (project)

    2023 – 2027 (expected)
  • Diploma in Artificial Intelligence
    Information Technology Institute (ITI)

    9-month programme in partnership with EPITA, Paris

    2021 – 2022
  • B.Eng. in Structural Engineering
    Tanta University, Faculty of Engineering
    2015 – 2020

Certifications

AWS Certified Machine Learning – Specialty
Amazon Web Services · Earned 2022
  • Data Scientist with Python
    DataCamp
    Oct 2021
  • Data Analyst with SQL Server
    DataCamp
    May 2022
  • Introduction to Tableau
    365 Data Science
  • AWS Academy Cloud Foundations
    AWS Academy

Honours

  • Best Engineer of the Year
    Empoweromics

    Awarded at Empoweromics for innovation and excellence in AI engineering.

09 Contact

Let's talk

Hiring for an AI, LLM or MLOps role, or planning a project? WhatsApp and email are the quickest ways to reach me. The form sends straight to my inbox. Or first.

It's in Cairo right now.

Or send a message

I reply from hatem@mohamedhatem.dev.

I'll reply to this address.

10 to 5,000 characters.

This form is protected by Cloudflare Turnstile.