Omar El Mountassir · Agentic Systems Architect · Marrakesh

Autonomous AI operations for SMBs & startups.

I design and build multi-agent systems that execute end-to-end — inside guardrails, with human checkpoints. Content, research, operations, software builds. My own business already runs on this stack, every day.

5+agents in daily production
1stinternational client — paid & settled
40+reusable agent skills
24/7governed execution, audited

Selected work

Each engagement, told the way it happened: the problem, my role, one decision that shaped the build, and the outcome. Statuses are real.

Cash Depot

Micro-lending · Montreal, Canada · web + programmatic video
Client · concluded
Problem

A Canadian micro-lending company needed a credible web presence and short-form video ads for social acquisition.

Role

Built the marketing website demo and the video production pipeline — research, brand extraction, programmatic rendering.

Decision

Programmatic video (Remotion, React) over template tools: every frame data-driven, versioned, brand-consistent at scale.

Outcome

Pitch site delivered — still online. First paid engagement (three TikTok-format video ads) delivered and settled.

See the live demo →

Guesthouse OS

Hospitality · Morocco · multi-agent operations
Pilot · concluded
Problem

A Moroccan guesthouse needed 24/7 operational coverage — bookings, guest communication, channel management — with minimal staff.

Role

Architected the system end-to-end: five specialized agents, defined workflows, single source of truth, PMS integration (HotelRunner API).

Decision

Governance-first: agents operate inside explicit guardrails, and sensitive actions require human validation — by design, not as an afterthought.

Outcome

Pilot delivered end-to-end. The system survives as a reusable template for the hospitality industry.

The Agentic Factory

My own operations · research, content, software
Live · daily
Problem

A one-person business cannot scale content, research and software builds by adding headcount.

Role

Built the factory: a durable daily-brief agent, a mechanical software factory, an automated content line with QA verification, unified memory across six agent harnesses.

Decision

Success is never the default: every build phase is born “failed” and earns acceptance through mechanical gates and recorded evidence.

Outcome

This stack runs my operations today — including the site you are reading.

Nova

Co-strategist platform · human ⇄ agent shared surface
Designed · backend live
Problem

A CEO and AI agents working together need one shared surface for state, decisions and tasks — not a dozen scattered chats.

Role

Designed the full platform: command palette, kanban, decision log, live agent status — complete design spec on a FastAPI + SQLite backend.

Decision

A cockpit, not a chatbot: the human stays the strategist; agents are a visible, auditable execution layer.

Outcome

Backend in daily production use via CLI. Frontend spec ready for implementation.

How we'd work

Three steps, plain words. The method is the same one that runs my own operations.

01

Audit

We map where your team burns hours. You get a prioritized list of automatable operations — with the numbers.

02

Build

Gated build waves. Every deliverable passes mechanical verification before you see it. You approve what matters; agents do the rest.

03

Operate

Systems run 24/7 inside guardrails, with human checkpoints on sensitive actions. You get an audit trail, not a black box.

About

Twenty-plus years in tech — from professional poker to agentic systems.

I spent eleven years as a professional poker player: decisions under uncertainty, strict capital discipline, pattern recognition from incomplete information. Same skills, new substrate.

In early 2026 I lost a five-figure web contract — not on skill, but on process: slow responses, no reusable infrastructure. So I built the infrastructure. Today my content production, market watching, software builds and invoicing run on governed AI agents I designed myself.

That stack is what I now build for clients.

What's eating your team's hours?

One email. Describe the repetitive process that frustrates you most — you'll get an honest take on whether, and how, it can be automated. No sales theater.

omar@el-mountassir.com

Prefer LinkedIn? linkedin.com/in/omar-el-mountassir