Available for consulting · Abu Dhabi / Remote

AI Engineer.
Systems
Architect.

Building AI-native infrastructure, LLM agent systems, and production-grade platforms. Currently within the G42 / Presight AI ecosystem in Abu Dhabi.

// current focus
whoami
mark — AI Infra · LLM Systems
location
Abu Dhabi, UAE 🇦🇪
stack --current
Astro · Bun · Railway · Drizzle
Claude Code · LiteLLM · Langfuse
status
open to contracts
AI Agents LLM Infra Arabic NLP GEO/SEO RAG Voice AI
5+
Years in LLM systems
12+
Production AI agents
3
Languages shipped in
Tokens processed
01 / consulting

How I can
help your team.

AI Agent Architecture

Design and build multi-agent systems, tool-use pipelines, and autonomous workflows using Claude, OpenAI, or open-source LLMs. From POC to production-grade infrastructure.

From $180/hr · or project-based
🔍
LLM Infrastructure & RAG

Build retrieval-augmented generation pipelines, embedding stores, and hybrid search systems. Optimize for cost, latency, and accuracy at scale.

From $160/hr · or retainer
🌐
GEO / Arabic AI Strategy

Generative Engine Optimization for Arabic-first markets. Help your brand appear in AI-generated answers across ChatGPT, Perplexity, and Gemini in MENA.

From $140/hr · strategy packages avail.
02 / writing

Articles &
Perspectives.

Featured · AI Infrastructure
Why KV-Cache Optimization and Progressive Disclosure Are at War in Multi-Tenant LLM Apps

The tension between serving dynamic, role-aware content and keeping inference costs low isn't a framework problem — it's an architectural one. Here's how to resolve it.

Read article →
March 2025·12 min read·AI · Infrastructure
Machine Learning
AgentBox: Building a Marketplace for AI Agent Skills
Feb 2025 · 8 min
GEO
Arabic GEO: The $2B+ Gap Nobody is Filling
Jan 2025 · 10 min
Workflow
My Claude Code Workflow in 2025
Dec 2024 · 6 min
Investing
How I Think About AI Moats in a Commoditizing Model Market
Nov 2024 · 9 min
03 / projects

Selected
Work.

01
PythonLiteLLMLangfuseFastAPI
Multi-LLM Agentic Loop

Provider-agnostic agent framework with unified tool interfaces, cost tracking, and support for Claude, GPT-4, and Gemini in a single opinionated loop.

02
AWS ECSPostgreSQLContextual Retrieval
Document Processing at Scale

RAG pipeline using Anthropic's Contextual Retrieval method with batched chunk enhancement, reducing embedding costs by ~97% vs naive approaches.

03
LiveKitFastRTCRailway
Voice Agent Infrastructure

Production voice agent platform evaluation and deployment. Compared FastRTC vs LiveKit architectures, shipping a reference implementation on Railway.

04
TimescaleDBPlotly.jsRailway
Trading Strategy Monitor

Real-time trading signal monitoring system with time-series storage, client-side visualization, and automated alerting built on Railway PaaS.

Got a hard AI
problem to solve?
Available for contracts, advisory, and short-term engagements.
Get in touch →
AI-indexed · llms.txt enabled