Hamza AlSalamat

About

Curious by nature, hands-on by habit.

The subject changes more than I do. Lately it has been AI native software, with the marketing and engineering that sit around it. Just as often it is a mat, a saddle, or a bow. What runs underneath all of them stays the same: pay attention, put in the reps, and keep getting better.

The longer version →

Get in touch

No introduction needed.

A line or two is plenty. If there is something in it, I will write back.

Write to me See the work

Hamza AlSalamat, signed by hand. hmsbuilds.com

Projects: things I built, start to finish

Projects

Things I built, start to finish.

Open any one.

Case File 01

Nama Site Intelligence

The engine I am building right now. Point it at a piece of land and it works out what is worth building there: the rules, the numbers, the recommendation. Minutes instead of months. Riyadh first.

Under the hood it is five agents in a line. One gathers the data and owns the archive, one interprets the zoning, one runs the numbers, one writes the deal sheet, and an orchestrator moves the work between them. They hand off through a single typed contract, so every stage knows exactly what it was given and what it owes the next one.

The discipline is in what it refuses to do. Every source row is kept exactly as it arrived, stamped with the hash and vintage of the file it came from, so any number on the finished sheet traces back to the row that produced it. Arabic is normalized before anything is matched, because a parcel written four different ways is still one parcel. And when a query only resolves as far as the district, that is not allowed to underwrite a specific plot. The engine returns nothing and says why. It is slower to build this way. It is the only version worth trusting.

  • Role Founder
  • Build Five agents, one shared contract
  • Rule Return nothing before returning a guess
  • Status In active development, Riyadh first
nama-solutions.dev soon
Dimensioned CAD drawing of the tubular steel chassis
Striking an arc on the chassis, sparks coming off the weld
Tacking steel box section at the outdoor bench
An instructor leaning in over a weld while the team watches
Disc brake, red caliper and steering knuckle at the front wheel
Night session, the rolling chassis under a cover
Grinding under the string lights, steering wheel already fitted
The first black body panel going onto the bare frame
The finished red and white car, dash and gauges in
1 / 10

drag, swipe, or use arrow keys

Case File 02

MISTI

An electric car, designed and overseen by MIT students and instructors, with our team on the tools. CAD drawings, then raw steel, then something that actually drives.

Global Teaching Labs is MIT's program, and it was hands on the whole way through. Their students and instructors led the design and the engineering, and we built it alongside them. The gallery walks the arc in order: the chassis drawing, arcs struck and tacked, an instructor leaning in over a weld, suspension and disc brakes, then the finished red and white body. The lesson had little to do with cars. A drawing stays a drawing until somebody picks up a grinder. The video has sound, so turn it on.

  • Program MISTI, MIT Global Teaching Labs, 2026
  • Team MIT students and instructors, alongside ours
  • Result It drives · 40 km/h flat out
See the certifications →
Paperly's landing page, where the resume goes in
A finished portfolio site Paperly generated from a resume
1 / 3

drag, swipe, or use arrow keys

Case File 03

Paperly

One of my earlier builds, made while I was teaching myself what AI could really do. Drop in your resume, answer three quick questions, and it hands back a finished portfolio site.

I wanted to know how far AI could carry a messy, real world input and still come back with something personal. A resume is a good test: everyone's is different, and most of them bury the interesting part. The build taught me more than the result did, which is roughly the point of building early.

  • Role Solo build
  • Why To find the edges of what AI could do
  • In → Out A resume → a portfolio site

Case File 04

The Outreach Agent

An agent I built to write outreach that earns replies, then to teach itself to write it better.

It runs the loop on its own: find the right people, research them, verify the contact, and draft in a voice that reads human. Nothing sends until I approve it. After that it watches what happens, who opened, who replied, what actually worked, and feeds every outcome back into the copy. The interesting part is not that it writes. It is that it keeps a record of being wrong, and edits itself accordingly.

  • What Self-learning marketing and sales agent
  • Loop Research → Draft → Approve → Send → Learn
  • Edge Copy that improves on evidence, not opinion

Case File 05

This Site

The one you are reading. Written by hand in HTML, CSS and JavaScript, with nothing sitting between the source and the browser.

The borders are the part I care about most. The tatreez running along the top and bottom of every page is not an image. It is drawn at runtime, stitch by stitch, the same cross stitch my family's embroidery is built from. The seven point star is generated the same way. I wanted the heritage in the site to be made the way the real thing is made, one stitch at a time, rather than photographed and pasted in.

There is no framework here and nothing to compile. That was a choice, not a shortcut. A portfolio that takes a moment to load is arguing against itself, and every dependency I skipped is one I do not have to explain, update, or trust. It loads fast because there is almost nothing to load.

  • Stack HTML, CSS, JavaScript. Nothing else
  • Tatreez Drawn in canvas at runtime, not pictured
  • Why Performance is a feature
Where all of this comes from →

Hamza AlSalamat, signed by hand.

Certifications

Certifications

Earned, not claimed.

Open any one of them. No download, and no need to take my word for it.

See what they led to →

Hamza AlSalamat, signed by hand.

About Hamza AlSalamat

About

Everything here started as something I couldn't do yet.

The simplest way to describe me is that I have to find things out for myself. Being told has never been enough. I want to stay close to something long enough to learn how it actually behaves, rather than how it is described.

Right now most of that is directed at AI. I have been building with it since it arrived, and I learned it the way I learn anything: by working with it until it made sense. It is a material to me now. I know where it gives, and I know where it will not hold. Nama Site Intelligence grew out of that.

That instinct has never confined itself to a screen. It is what put me on a mat, where I still compete in kickboxing, jiu-jitsu and wrestling, and where I have taken both gold and silver.

Competition taught me the lesson that carries into everything else. It tells you immediately, and in front of people, exactly what you skipped. Eventually you stop skipping.

Nobody starts perfect.

Outside of that I am usually outdoors: riding, or up in the mountains. I would rather get chased by an animal than read about one. The same instinct, somewhere quieter.

What is on the rest of this site is what the habit has produced so far. The habit is the part worth knowing.

The whole of it

Everything I keep in play.

One sky. Nothing in it stands alone.

Right now

  • Building Nama toward its launch in Riyadh.
  • Learning whatever the next build turns out to need.
  • On the mat most weeks, and in the mountains when I can get there.
Projects See what came of it → Certifications Earned, not claimed → Contact Straight to me →

Hamza AlSalamat, signed by hand.

Contact Hamza AlSalamat

Contact

Open line.

Email, LinkedIn, GitHub. No forms and nobody in between.

Email HMSalamat15@gmail.com

Hamza AlSalamat, signed by hand.

Certificate Open full PDF ↗