Event-sourced spine
Postgres with pgvector underneath. Every state change is an event, so the system can be replayed, audited and reasoned about rather than just queried.
Systems
I got tired of solving the same problem twice.
So I learned to build. These are the systems that came out of twelve years of watching what actually breaks.
Flagship
Auto Pilot Events OS
An AI operating system for running an experiential events company. Built solo in 28 days, from 15 June to 12 July 2026. In production usage since the day it shipped.
95Registered tools
across 29 families
across 29 families
7+1Specialist agents
plus a director
plus a director
753Tests
green
green
28Days,
solo build
solo build
No client ROI figures yet. When there are, they will appear here with a source.
Under the hood
What it is actually made of.
Built for an established UAE experiential events group operating across the GCC.
Seven agents and a director
Specialist agents for the distinct jobs, with a director that routes work between them instead of one model trying to hold everything at once.
Twenty three page application
A working Next.js front end, not an API and a promise. Roughly 25k lines of Python and 14k of TypeScript behind it.
Four cost modes
Runs from zero dollars a month up to roughly three hundred, depending on which models are enabled. The cheap mode is genuinely usable, which was the point.
The model
Pricing, cross-validated on real data.
The part I am most willing to be questioned on, because the numbers came out of a validation run rather than a pitch deck.
| Measure | Result |
|---|---|
| Real price points used | 1,136 |
| Baseline mean absolute error | 41,432 |
| Model mean absolute error | 13,927 |
| CatBoost promoted over baseline | +7.6% and +8.1% |
Twelve years of knowing what breaks is the actual moat. The code is the easy half.
Ibrahim Muhammad Naeem
Also running
The rest of the stack.
Brain
A personal AI operating system. Domains, persistent memory, scheduled briefs and a runner that does not sleep. I use it every day, which is the only review that matters.
Data science
Kaggle Expert with six published notebooks. Breast cancer detection at 98%, churn prediction at 86%. Python, SQL, Power BI and Tableau.
Verified-facts pipeline
One source of record behind every figure published under my name. If a number is not in it, it does not go on a page. That rule is why this site is short on superlatives.
Build with me
What is the task you keep redoing?
That is usually the one worth automating first. Tell me what it is and I will tell you honestly whether it is worth building.