Founder · Strategist · Agentic architect

I build companies
with AI agents
and no engineering team.
Then I write about
what I learn.

I'm Muhanad — a non-technical founder who built Steller, a live B2B gift card platform, through AI agent orchestration. Before that, a decade of business development in one of the world's hardest markets.

Steller — built with AI agents
$672
Total cost to rebuild a live B2B platform from scratch

27
Days from broken microservices to agent-ready monolith

<60s
Time to first partner value on the new onboarding flow

0
Full-time developers required to operate Steller today
Who I help

Three kinds of founder
I work with.

I'm available for advisory, consulting, and fractional product work. Here's where I add the most value — and what you walk away with.

01
For · Non-technical founders
You have the idea and the market. You need to build without a dev team you can't afford yet.
I'll show you how to design your system for AI agents from day one — the architecture decisions, the documentation layer, the protocol that lets an AI operate your product reliably. You get a working system without a payroll.
Outcome: agent-operable product at startup cost
02
For · Technical founders & CTOs
Your system works. Your AI agents keep breaking things they couldn't see.
I've lived this exact problem — and solved it by learning that microservices are architecturally hostile to AI agents. I'll audit your system for context visibility, identify the invisible dependencies, and help you restructure for agentic reliability.
Outcome: architecture that AI agents can actually operate
03
For · Product teams
You have a developer API. Partners are signing up and churning before they place a first order.
I designed Steller's 60-second path to value — the PLG 2.0 onboarding flow that shifts the partner from builder to reviewer and gets them to a working integration before they've read a word of documentation. I'll apply this thinking to your product.
Outcome: faster TTV, lower churn, more confident partners
The essay series

Writing that comes from
building, not theorising.

Every essay is grounded in something that actually happened — a decision made under pressure, a system that broke, an insight that only exists because we tried and failed first.

Operating notes

A second series, with its own loop. The three essays above stay the Steller Story. These ten follow the file an agent reads first, then the partner path, and end on the published cost.

Note 01 · Operating notes
April 16, 2026 · 9 min read
What Belongs in AGENTS.md
Router, hard rules, and the locks an agent is not allowed to relax.
AgenticRead →
Note 02 · Operating notes
May 2, 2026 · 7 min read
The Agent Handover Protocol
What the hardening week named, and what can still be quoted.
AgenticRead →
Note 03 · Operating notes
May 18, 2026 · 7 min read
What Zero Full-Time Developers Actually Requires
The architecture, the docs, and the human approval that remain.
AgenticRead →
Note 04 · Operating notes
June 3, 2026 · 6 min read
When Fat Controllers Stop Working
The exit conditions the rubble essay already stated.
AgenticRead →
Note 05 · Operating notes
June 19, 2026 · 7 min read
Why Explicit Jobs Beat Invisible Events for AI
Named Hangfire work, and the Bamboo locks around it.
AgenticRead →
Note 06 · Operating notes
July 5, 2026 · 7 min read
The Failure-First Onboarding
The three failures both essays surface, cited in each essay’s order.
Partnership / PLGRead →
Note 07 · Operating notes
July 21, 2026 · 7 min read
202 Accepted Is a Product Decision
Accepted is the value signal. The PIN comes later.
Partnership / PLGRead →
Note 08 · Operating notes
August 6, 2026 · 7 min read
Idempotency for Gift Card APIs
referenceId for partners, and a separate vendor Idempotency-Key.
Partnership / PLGRead →
Note 09 · Operating notes
August 22, 2026 · 7 min read
Doing the Customer’s Work Before They Ask
The published career through-line, and the 60-second path it became.
Partnership / PLGRead →
Note 10 · Operating notes
September 7, 2026 · 7 min read
The Economics of Agent-Operated Software
The published spend, labeled as published, and what was not disclosed.
AgenticRead →
The methodology

Agentic architecture —
the principles I work by.

These aren't abstract ideas. They're the specific decisions I made while rebuilding Steller — and the ones I help other founders make.

01
Optimise for the context window, not the service boundary
Architecture decisions must first answer: can an AI hold the entire relevant flow in one context window? If not, the design is wrong — regardless of its engineering merit.
Context visibility over distribution
02
Keep the golden flow in one traceable path
The sequence that moves revenue must be readable top to bottom without crossing a service boundary. If an AI can't trace it in one sitting, redesign it.
One flow · One file · No invisible breaks
03
Write AGENTS.md before you write any code
The file that tells an agent where to start is worth more than any single code file. It cuts session orientation from 20 minutes to 2. Written once, paid forward forever.
Docs compound like interest
04
Make the wrong move require explicit confirmation
Guardrails, confirmation gates, protected systems. An agent that cannot accidentally destroy a production system is worth ten agents that might.
Safety is architecture, not comment
Background

Two careers that shouldn't
go together — and do.

Sudan
Zain Sudan
Radio tower certification.
Operations
Emirates Driving Company
Operations redesign.
Partnerships
Rittal Technology
Partnerships.
Company
Steller
Platform architecture, onboarding, and company building. A live B2B gift card platform operated with AI agents and no full-time developers.
Emerging market business development

A decade of deals in markets defined by risk, scarce information, and internal politics — Sudan especially. Winning often meant doing the customer's internal work, clearing blockers, and training teams on products they had never seen. That is where PLG 2.0 came from: do the first part of the job before they have to ask.

2014Business development, East Africa and MENA markets
2018Complex B2B sales, emerging market partnerships
2022Identified the B2B gift card middleware gap
2024Founded Steller — first attempt with outsourced team
Agentic product development

Building Steller without a team showed that architecture decides whether an AI agent can operate a system at all. The result is a documented method, the essays on this site, and a live B2B platform run by Qwen Coder with no full-time developers.

Late '25Dev team departure — pivot to full AI development
Feb '2627-day rebuild — microservices → agentic monolith
Mar '26Bede (Zain) first live integration — system handed to Qwen
NowWriting, advising, building the methodology
"In Sudan, you learn to win contracts by doing the customer's internal work for them — navigating their politics, training their teams, removing every reason to say no. We didn't know we were inventing a product philosophy. We were just trying to close deals."
On the business instinct that became PLG 2.0
Contact

Let's work together.
Or just talk.

Advisory, fractional product, and consulting on agentic architecture and PLG. If you're figuring this out for the first time, I'm happy to talk.

Steller — the product
developers.steler.org
B2B gift card API. If you're a platform looking for a reliable gift card integration, start here.
Visit developer hub →
Work with me directly
Muhanad Abdelrahim
Advisory, consulting, fractional product. Agentic architecture audits. PLG onboarding design. Emerging market strategy.
hello@steler.org → LinkedIn →