Not a portfolio. The system behind the work, the evidence it produced, and what each part of it does for a business.
Not estimated, not rounded up. The count excludes anything downloaded rather than written — no libraries, no dependency folders, no duplicated working copies — and identical files are counted once.
An earlier draft of this slide said 5,336 Python files. That number counted vendored dependencies. Corrected, the real figure is 853. It is a smaller number and it is the true one, and you should assume everything else here was checked the same way.
134 project directories. 91 under version control. 1,812 commits. 277,314 lines. 1,866 built pages, 679 PDFs, 2,213 written records.
Five layers. Each one exists because something went wrong without it.
Work passes down through all five. Nothing skips a layer, including work done at two in the morning.
The rule that produced them: no project computes its own physics. If the engine lacks something, extend the engine, then import it.
A specification is prose. A joiner reads sentences, and sentences pass every numeric check.
One build carried a retired ceiling construction — the wrong slat width, the wrong slot, the wrong total depth — while four separate numeric gates passed green. A contractor would have been handed a ceiling that does not fit the void that exists.
There is now a gate on the prose, and it was tested by planting the wrong words and requiring it to fail.
A treatment schedule that added absorption instead of substituting it for the wall it covers — optimistic by up to 30% as coverage rose.
A 3D view that drew 2.8 times the absorption the report charged for. The picture is what a client believes.
A drawing with a chair standing inside a structural column — every numeric gate passed, because the numbers were right about a room that did not exist.
All three are in a written register with what each one cost. The register is the reason they do not recur.
One system, built once, that now handles every job of its kind.
Acoustic work needs reverberation, intelligibility, room modes and loudspeaker coverage computed for every room. Across projects there were six separate implementations of that arithmetic, and each project had picked whichever one it inherited.
On identical inputs they disagreed by up to 62.6%. Four of the six applied no air absorption at all; a fifth used coefficients about four times too aggressive. A published figure on a live client site was wrong by exactly 10 dB — it compared the combined output of many loudspeakers against the reverberant field of one.
None of those errors threw an error. Every one returned a plausible number. That is the failure mode all of this exists to catch.
Plus a Python port of the same physics, held to the JavaScript within one part in a billion by automated parity tests — because two engines that disagree are worse than one.
And a decision log now past its hundredth entry, each recording what changed, why, and what it superseded.
The engine was built for one project. It has since carried mosques, a broadcast studio complex, a conference hall, a masjid basement, a recording room and a corporate floor — in two languages, with drawings, bills of quantity and commissioning criteria.
The marginal cost of the next room is a fraction of the first, and the quality goes up rather than down, because every defect found anywhere becomes a test everywhere.
That is the pattern worth looking at, and it has nothing to do with acoustics. Any problem a business solves repeatedly can be turned from a service into a system — and the same five layers apply to invoices, approvals, compliance records and reporting.
Fifteen systems, grouped by what they do. Every claim countable from the disk.
A room-acoustics engine written from the physics up: reverberation by four analytic models plus a geometric image-source solver, room modes, speech intelligibility to IEC 60268-16, air absorption to ISO 9613-1, loudspeaker coverage from measured manufacturer balloon data, absorber design from porous-medium theory, and a treatment solver that sizes material per frequency band.
| Detail | |
|---|---|
| Evidence on the disk | 250 commits · 29,929 lines · 38 exported functions in the core · 11 extension modules · a Python port held to bit-parity against the JavaScript · 24 recorded adversarial reviews |
| What it does for a business | Any room, any building: a study a contractor can build from, and a specification that survives a consultant's review. |
| Indicative range | $4,000–$18,000 per room depending on evidence required — a range for orientation only; every real number comes after a scope conversation |
The same physics in Python so it can run inside report builders and data pipelines, held to the JavaScript to within 1e-9 by automated parity tests. Two engines that disagree are worse than one; the tests are what stop that happening.
| Detail | |
|---|---|
| Evidence on the disk | 28 commits · 2,311 lines · parity tests on air absorption, intelligibility and object handling |
| What it does for a business | Reporting that computes rather than transcribes — every figure in a document traceable to the code that produced it. |
| Indicative range | included in a study; licensed separately for a firm — a range for orientation only; every real number comes after a scope conversation |
Before anything ships, the substantive claims are put to six independent models across OpenAI, Google, xAI, Mistral and Meta, in parallel, each writing its own receipt. Disagreement is resolved by running the code or citing the source — never by counting votes. Two of the seats can read files and run programs to check themselves.
| Detail | |
|---|---|
| Evidence on the disk | a command-line orchestrator with a seat roster, dispatch, status and an append-only round ledger; every round stored with the brief's hash and every seat's raw reply, including the ones that failed and why |
| What it does for a business | A second, third and fourth opinion on a decision before it is made, at no marginal cost — and a written record of who said what. |
| Indicative range | built into every engagement — a range for orientation only; every real number comes after a scope conversation |
Executable checks that run before a deliverable is built: does the specification prose match the locked design, do the quantities in the drawing match the schedule, is every figure traceable, does the Arabic contain the defects Arabic acquires when it is translated rather than written. A gate that has never been made to fail proves nothing, so each one is tested by planting the defect it is meant to catch.
| Detail | |
|---|---|
| Evidence on the disk | gate suites in the engine and in four client projects; a git pre-commit hook that blocks a build carrying a retired specification; a client-safety audit that scans in both Latin and Arabic numerals |
| What it does for a business | The difference between a document that looks right and one that has been proven right. |
| Indicative range | built into every engagement — a range for orientation only; every real number comes after a scope conversation |
Every opportunity, quoted or declined, in one file with its own identifier; a billing registry that syncs from it every three hours and never overwrites a status a human has advanced; and an inbox watcher that files pricing-relevant email as draft rows needing review rather than guessing.
| Detail | |
|---|---|
| Evidence on the disk | clients/NHJ: 229 commits · 94 built pages · 62 documents · live at billing.obaisukar.com |
| What it does for a business | You stop losing track of what was quoted, to whom, at what price, and whether it was ever paid. |
| Indicative range | $6,000–$14,000 to build, then hosting — a range for orientation only; every real number comes after a scope conversation |
A post-approval checkpoint with an append-only ledger: work is scored against a written standard before it is published, and the record of who approved what cannot be edited afterwards.
| Detail | |
|---|---|
| Evidence on the disk | live at stamp.obaisukar.com, in pilot |
| What it does for a business | Nothing leaves the building unreviewed, and you can prove it a year later. |
| Indicative range | $5,000–$12,000 — a range for orientation only; every real number comes after a scope conversation |
A single index of every project with its infrastructure facts written by one feeder every four hours and its live status written by another every thirty minutes, reconciled so neither silently overwrites the other.
| Detail | |
|---|---|
| Evidence on the disk | 73 projects indexed; two independent feeders; a bridge that paginates so long records are not silently truncated |
| What it does for a business | An owner with several businesses stops asking where things stand and starts reading it. |
| Indicative range | $8,000–$20,000 — a range for orientation only; every real number comes after a scope conversation |
Bilingual acoustic and audio-system studies with drawings, bills of quantity, commissioning criteria and vendor handoff packs — built so a contractor can price and build from them without asking a question.
| Detail | |
|---|---|
| Evidence on the disk | Al-Furqan 153c · Al-Iman 140c · Mazzeh · Refayee 67c · Haitham Halabi 128c · 76 issued documents across them |
| What it does for a business | A study you can hand to a builder, in the builder's language, that does not come back with questions. |
| Indicative range | $4,000–$25,000 — a range for orientation only; every real number comes after a scope conversation |
Website, brand system, media production, campaign material and an acoustic programme for a technology campus — the largest single body of work on the disk.
| Detail | |
|---|---|
| Evidence on the disk | 139 commits · 80,287 lines · 788 built pages · 185 documents |
| What it does for a business | One supplier for the whole estate instead of five who blame each other. |
| Indicative range | scoped per programme — a range for orientation only; every real number comes after a scope conversation |
Complete brand systems built from a strategic platform rather than a mood board: the name, the trademark screen, the marks in vector, the identity in use across the surfaces it will actually live on, and the campaign lines that carry it.
| Detail | |
|---|---|
| Evidence on the disk | Madafa — name, arch mark, 28 art-direction plates, a 28-slide bilingual deck, brand-in-use mockups, 24 commits · Sukar Studios · SPC sonic identity · Mealy · RMZ |
| What it does for a business | A brand with a reason behind every choice, and the files to deploy it the same week. |
| Indicative range | $7,500–$30,000 — a range for orientation only; every real number comes after a scope conversation |
Design and build, deployed on infrastructure I run, with certificates, security guards and version control — not handed to a hosting company and forgotten.
| Detail | |
|---|---|
| Evidence on the disk | 83 sites served from one machine; deployments verified by comparing the local file, the server file and a cache-busted fetch before anything is called done |
| What it does for a business | A site that ships in weeks and stays up, with someone accountable for it. |
| Indicative range | $4,000–$22,000 — a range for orientation only; every real number comes after a scope conversation |
Arabic composed from the fact rather than from the English sentence, with a grep-able gate that catches the constructions translation produces — the propped-up verbal noun, the calqued aphorism, the dead metaphor — plus Arabic-Indic digits and Levantine months as house style.
| Detail | |
|---|---|
| Evidence on the disk | a proofreading standard applied across every Arabic deliverable, and a detector with a positive control, because a checker that silently matches nothing passes everything |
| What it does for a business | Arabic that a ministry reads as native. Most vendors cannot supply this at any price. |
| Indicative range | included; $1,200–$4,000 standalone — a range for orientation only; every real number comes after a scope conversation |
Original composition and arrangement produced as notes, harmony and counterpoint in code, then rendered through sampled instruments to MIDI, stems and a mixdown. Not a generative audio toy — it writes the score.
| Detail | |
|---|---|
| Evidence on the disk | 58 python modules · 10,732 lines · used for client themes and sonic identity |
| What it does for a business | A theme, a sonic identity or a cue that is yours outright, revisable to the note. |
| Indicative range | $2,500–$15,000 — a range for orientation only; every real number comes after a scope conversation |
Direction, edit, sound design and delivery — 25 years of audio engineering behind the mix.
| Detail | |
|---|---|
| Evidence on the disk | mal-launch-film 34c · TechTown showcase film · Komitas 10c · Karazah, an Arabic channel at scale |
| What it does for a business | Film that sounds like it cost more than it did. |
| Indicative range | $3,000–$40,000 — a range for orientation only; every real number comes after a scope conversation |
Coverage, intelligibility and level uniformity computed from measured loudspeaker data — then specified, procured, commissioned and measured against the prediction.
| Detail | |
|---|---|
| Evidence on the disk | the Umayyad Mosque in Damascus under a Ministry commission over Dante networking · Masjid Omar · AlUla · Refayee basement · TechTown event area · Syria MOI phase two |
| What it does for a business | A system that is intelligible in the last row, proven before it is bought. |
| Indicative range | $3,500–$20,000 for design; installation separate — a range for orientation only; every real number comes after a scope conversation |
Ranges are for orientation. Nothing here is a quotation, and every real figure follows a scope conversation — a price given before the brief is read is a guess with a dollar sign on it.
| Capability | What you get | Indicative |
|---|---|---|
| Engineering software | SUKAR | $4,000–$18,000 per room depending on evidence required |
| Engineering software | SAE | included in a study; licensed separately for a firm |
| Quality control | The council | built into every engagement |
| Quality control | Gates | built into every engagement |
| Business operations | Deal-flow ledger and billing platform | $6,000–$14,000 to build, then hosting |
| Business operations | Publish Gate | $5,000–$12,000 |
| Business operations | Mission control | $8,000–$20,000 |
| Documents that carry a specification | Contractor-executable engineering studies | $4,000–$25,000 |
| Documents that carry a specification | TechTown Damascus | scoped per programme |
| Brand and identity | Naming, identity and campaign platforms | $7,500–$30,000 |
| Brand and identity | Websites, built and hosted | $4,000–$22,000 |
| Brand and identity | Institutional Arabic that was written, not translated | included; $1,200–$4,000 standalone |
| Sound and media | score-craft | $2,500–$15,000 |
| Sound and media | Film, campaign and launch video | $3,000–$40,000 |
| Sound and media | Public-address and AV system design | $3,500–$20,000 for design; installation separate |
And what it needs from you.
Access to the real numbers, not a summary of them. The whole method rests on measuring rather than estimating.
One decision-maker per question, and a date.
Tolerance for being told the unwelcome version. A model that only produces good news is a sales document.
Invent a figure to fill a gap. An empty line that names who to ask is worth more than a filled one that says nothing.
Promise an outcome it does not control. Booking volumes, occupancy, revenue — those are your P&L, not my deliverable.
Quote before reading the brief.
Hide a mistake. Corrections are written down with what they superseded, including the ones found the same night.
This searches the catalogue on this page. It runs with no network, it cannot invent an answer, and if it has nothing it says so. Type a business problem or a word — invoice, arabic, website, music, compliance, approval, room.
One person, with engines that compute instead of estimating, gates that refuse to ship a wrong document, six models attacking the work before anyone else sees it, a record that cannot be quietly edited, and delivery that is proven rather than assumed.
1,812 commits. 277,314 lines. 1,866 documents. 83 live sites. One machine.
Which means the useful question is not what is on this list. It is what you need that is not on it yet.