Home/Services

Use cases

The problems I get called for, what I build for each one, and the proof that I've built it before.

Work asFull-time · contract · studio · workshops
CoreAI products, agents & enablement
LanguagesArabic & English
Based inAmman, Jordan · GMT+3
Book a 30-min call Hiring? Get my CV
01 / If this sounds familiar

Five problems
I solve.

Use case 01

Invisible agents

"We have agents in production and nobody can tell which ones are working, what they cost, or why one broke."

What I build
A control plane: heartbeats and health scores for every agent, cost per agent and per app, remote pause, model swap and prompt push, versioned prompts with rollback.
Proof
Agent Control Panel · 350+ agents across 35+ apps
Use case 02

Keys everywhere

"Every app holds its own database credentials, schema changes break things silently, and onboarding an app takes days."

What I build
One gateway with table- and column-level permissions, request logs, friendly errors and rate limits, plus AI-generated integration clients for each new app.
Proof
MCP Hub · onboarding went from days to minutes
Use case 03

Wrong context

"Our agents see too much, too little or the wrong thing, and we can't explain why an answer came out the way it did."

What I build
A context layer that ranks sources by value per token, keeps untrusted text out of instructions, and records every decision so it can be replayed and audited.
Proof
Context Engine · 9-stage pipeline, byte-for-byte replay
Use case 04

AI-curious team

"Our engineers want to use Claude Code, skills and MCP properly, not just paste into a chat window."

What I do
Hands-on workshops and setup sprints: Claude Code, writing skills, wiring MCP, building agents your team can trust, in Arabic or English.
Proof
190+ people across four workshops, plus five diploma cohorts
Use case 05

An MVP that holds

"We need a product built, our budget is real but small, and we can't afford something that falls over."

What we build
With SKAI Group, the studio I co-founded: AI systems, web, mobile and SaaS. We scale the scope to your budget. Never the quality.
Proof
4 clients closed in 2 weeks · Al Delma ISO training platform
02 / How a project runs

Four steps.
No mystery.

Step 1 of 4

Map

Before I suggest anything, I measure what exists: a snapshot query of your data, a list of every app and agent, and where they actually break. Arguments start from numbers, not adjectives.

You keep
  • A written picture of your system, with counts
  • Failure classes ranked by impact
  • A short list of what not to touch
03 / The Claude ecosystem

Taught by
someone who
ships with it.

Claude Code is my daily driver. I've built an MCP gateway, a skills CMS that A/B tests skills against real agents, and a control plane for 350+ agents.

That's the difference between a workshop that demos features and one that leaves your team with a working setup.

04 / Workshops

Admit one
team.

Hands-on workshop · team admission Claude Code
Ecosystem
FormatHalf day or full day
SeatsYour team
LanguageArabic or English
WhereOn-site, Amman

Also available: Build Your Own AI Agent, Build Your Own AI Chatbot, and AI for Business.

Stub · keep this Book
it
Full-time

AI engineering roles across the stack: agents, RAG and data, and the product around them.

Open now
Contract

Scoped builds and audits: a gateway, a control plane, a context layer, or a database health check.

Hourly or per project
Studio

Products and MVPs through SKAI Group, the AI-native studio I co-founded in Amman.

Scope scaled to budget
Workshops

Training for engineering teams and companies adopting the Claude toolkit.

Half or full day

Not sure which one fits? Book 30 minutes. We'll work out what you need, and I'll tell you honestly if I'm not the right person for it.

Book a 30-min call
05 / Questions people ask
01Do you only work with Claude?

No. My production systems run Claude, OpenAI, Gemini, Grok, Groq and OpenRouter behind one gateway with fallback chains. Claude is where I go deepest, and the toolkit I teach.

02Can you work inside our existing stack?

Yes. My default is TypeScript, Node, PostgreSQL or Supabase and React, with Python for ML. Either way, I start by mapping what already exists before I suggest changing anything.

03What does an audit actually look like?

A snapshot query of what's really in your database, then failure classes with counts and a fix plan for each. The last one I ran found 5 failure classes across 9 apps and took signal throughput from 78 to 864.

04Can you show us code?

Systems I built for employers belong to them, so for those I walk you through architecture and decisions. My own projects I can show directly.

05Arabic or English?

Both. I write, teach and present in either.

Contact

Let's build

Tell me which use case is yours. One email is enough to start.

Book a 30-min call Hire on Upwork Download CVTrainer CV