Mobile Engineer · AI Systems · Dubai

The next app won't be tapped through.
It'll be understood.

I'm Kirollos — 11+ years building native iOS at scale. Now I'm building the infrastructure for an AI-first world, where agents act on our behalf and software renders itself, on-brand, for each person.

Where I'm going → Work with me
Intent Agent MCP Brand + data Rendered UI
scroll to explore ↓
Who I am

I ship software to millions —
and I'm rebuilding how it gets made.

Kirollos Magedy

For over a decade I've shipped native iOS: a 1M+ user app at 99.99% crash-free, senior mobile work across e-commerce, and an internal AI platform I initiated and led — on-device and server LLMs with multi-agent orchestration, running in production.

Egyptian, based in Dubai, and happiest right at the edge of what's just become possible. I build in the open, and I move fast toward the future I believe in.

↓  View my résumé
What I bring

Depth in mobile. Fluency in AI.
A systems mind.

// native iOS

SwiftSwiftUIUIKitCombineasync/awaitSPMCore Data

// AI & agents

Claude & OpenAI APIsmulti-agent orchestrationon-device MLRAG / contextMCP

// architecture & delivery

system designMVVM / CoordinatorCI/CDautomated testingperformance at scale

// leadership

initiating platformstechnical leadershipmentoringcross-functional delivery
What I believe

My read on where this goes.

01

The interface becomes generated, not shipped.

Static screens give way to UI an agent composes at request time — grounded in real data and a real brand, so it's on-brand without being hand-built.

02

MCP is the backend for AI-native products.

Every system exposes its data, its actions, and its design language as agent-friendly tools — not opaque dashboards or frozen templates.

03

Brand becomes a system an agent can render.

Encode a brand as machine-readable tokens — color, type, spacing, voice — and an agent can render a thousand true variations instead of one static theme.

The plan

How I'm going to change
how software gets built.

01

Collapse the loop

Requirement → user stories → UI → implementation → automated tests, in one continuous pass. My Create & Analyze plugins already do this.

02

Ground the AI

The Vault: an index of a codebase, its components and APIs, so an agent knows exactly what it touches and what to reuse. Stories become executable plans.

03

Render the brand

An MCP that carries a brand's design language, so agents compose on-brand experiences per person — the storefront as something generated, not frozen.

04

Open it up

MCP as the connective layer for AI-first products — starting with commerce and Shopify, then everywhere software meets an agent.

Latest project

Create User Story —
grounded by the Vault.

// AI plugin · agentic SDLC

A user story that already knows your code.

A user story on its own can't drive development — it doesn't know the codebase. So I built the Vault: an index of a codebase, its UI components, and its APIs. The Create User Story plugin writes against it, so every story already points to the exact modules it touches and the components to reuse — and generates UI from the product's real look and feel.

inputRefined requirement
→
the vaultCodebase · UI components · APIs, indexed
→
outputStory with affected modules, reusable components & generated UI

// problem

Stories are wishes — they don't map to real code, so they can't lead implementation.

// approach

Index the code into the Vault; write stories against it so they reference what actually exists.

// result

Stories become executable plans — and new engineers ship in week one, not quarter one.

1 loop
requirement → tested code
weeks → days
onboarding ramp
grounded
every story tied to real code
What I'm working on

Workflows for software engineering.

2026 · 08

Create & Analyze User Story, live

Plugins that carry a requirement all the way to implementation and automated tests — in one continuous loop.

2026 · 07

Requirement → production, one pass

Collapsing the delivery cycle so a refined requirement flows straight to UI, code, and tests — nothing lost in the handoffs.

2026 · 06

The Vault

An index of a codebase, its components and APIs, that grounds every generated story in real, reusable code.

What's next
New devices are coming — and they'll be AI-first.

The phone was built for taps. The next generation of devices will be built for agents — ambient, intent-driven, rendering experiences on demand. Software will be written for the agent as much as for the human. That's the shift I'm preparing for, and building toward — one layer at a time.

Let's build

If you're building the AI-first future, let's talk.

Open to consulting on MCP & AI systems, and to senior roles building this next layer. Fastest ways to reach me:

Email LinkedIn GitHub Résumé