Mohamed Hatem
AI Team Lead Globant Multi-agent LLM systems, RAG, LLM serving
- mohamedhatem.dev
- hatem@mohamedhatem.dev
- +20 114 207 0946
- in/mohamed-hatem-6a5790173
- github.com/mhatem5351
- Cairo, Egypt · Remote
Summary
AI Team Lead with 7 years in data and AI, designing, shipping and running production LLM systems: multi-agent orchestration (LangGraph), RAG and hybrid search, self-hosted model serving (vLLM on NVIDIA A100) and LLM evaluation. Built a 14-agent assistant for ArabsStock and a 6-agent bilingual Arabic/English chatbot at Wakeb Data. AWS Certified Machine Learning – Specialty (2022).
Experience
- Leading AI engineering on the Saudi Tourism Authority (STA) project: an agentic LLM runtime with retrieval and evaluation for Noura, the visitor assistant, and a staff assistant, in Arabic and English.
- One LangGraph agent behind 96 tools (27 visible to visitors), with human approval before every action; Noura was accepted in the client's UAT in September 2026.
Client: Saudi Ministry of Defense (GADD)
- Architected and deployed a fully async 6-agent bilingual (Arabic/English) LangGraph chatbot on Qwen, served with vLLM on NVIDIA A100 GPUs.
- Owned model serving across multiple production GPU servers: vLLM configuration, GPU allocation, model versions, health monitoring and automated restarts.
- Built hybrid RAG (vector + BM25 with LLM reranking), node-level caching and Brave Search over MCP, shipped as Dockerised FastAPI services behind Nginx.
- Designed AIP semantic search v2, built a bilingual enterprise assistant for a Saudi government client with LLM evaluation, and moved key APIs from REST to gRPC.
- Architected and built the ArabsStock Assistant: 14 LangGraph agents for semantic search, AI image generation, database operations and content moderation.
- Built semantic search (OpenAI embeddings + Pinecone, automated batch ingestion) over millions of photos, videos and vector graphics, plus visual-similarity search.
- Shipped audio AI for a beta sound library (generation and search, instrument classification, BPM detection, watermarking) as containerised FastAPI services.
- Set up agent evaluation to track reliability, response accuracy and routing in production.
Software Engineer Tabaqat Feb 2023 – May 2024 Part-time · Remote (Saudi Arabia) - Designed and managed all data pipelines in Prefect, including ETL for GIS data from Saudi government sources.
- Queried Overture Maps with AWS Athena and DuckDB; deployed map serving with MapProxy, MBTileServer, MapLibre and Kepler.gl.
Data Scientist Empoweromics Best Engineer of the Year Jan 2022 – Mar 2023 Full-time · Cairo, Egypt · Case study - Unified data from 3,008+ real-estate developers, each with its own schema, to scale the company's real-time e-map.
- Built lead-lifecycle models (7 stages) for broker prioritisation, and text + image WhatsApp classification that routes messages to the right department.
- Built Smart Search (fuzzy lookup over misspelled or partial Arabic/English names) and a PyTorch model inferring gender from Arabic names.
- Designed ETL from Cosmos DB to SQL Server, scheduled with Azure Functions.
AI Research Assistant · The British University in Dubai · Umm Al-Qura University Mar – Nov 2022 Research assistant and AI lecturer for master's students in biomedical informatics at BUID; mentored master's research at BUID and Umm Al-Qura University.
AI Instructor · Military Technical College Aug 2022 Taught AI courses (ML fundamentals, deep learning, deployment) to engineers in the Egyptian Armed Forces.
Freelance Data Scientist · AestheticsPro (US healthcare software) Feb – Apr 2022 Scraped, cleaned and loaded healthcare data into SQL Server; built EDA and a stakeholder dashboard. Built ML models predicting patient retention, used to target retention offers.
SQL Developer · Misria Digital Payments (MDP) Jan 2020 – Jan 2021 Developed and optimised complex SQL stored procedures for production payment systems; query tuning and database administration.