Inside the product
The recruiter workspace, from the evidence dashboard to the human decision.
Captured by the repo’s own Playwright suite on a seeded local build. Candidates, scores and jobs are demo data.
Backend & Business Systems Engineer
I turn businesscomplexity intoproduction systems.I turn business complexityinto production systems.
Backend systems built around how real businesses operate.I design backend systems around how real businesses operate — workflows, permissions, data, APIs, integrations, reliability and applied AI.
Backend SystemsBusiness WorkflowsApplied AISyria / RemoteScroll
Businesses don't run on features. They run on operations.
Sales pipelines, commissions, bookings, approvals, stock, schedules. Every business has rules that decide who can do what, when, and with which data. That is where I work.
- 01Workflows that match how teams actually work
- 02Permissions enforced on the server, not the screen
- 03Data modeled so reports can be trusted
- 04AI that assists without taking the decision
Real-estate ERP · Arabic-first
Rakez ERP
An operating system for a real-estate business: sales, bookings, commissions and marketing spend in one platform.
Rakez ERPWhat the system handles
- BookingsReservations with a waiting list for contested units
- CommissionsMulti-party splits, approval chains and VAT
- PermissionsGranular roles, checked on the server
- Real-timeWebSocket notifications through Laravel Reverb
- AssistantOpenAI-backed answers in Arabic for the sales team
Laravel 12 · PostgreSQL · Redis · Reverb · Sanctum · Spatie Permission · OpenAI API
Read the Rakez case studyFrom a business rule to a production system.
One commission rule from Rakez, followed down through every layer it touches.
- 01The rule
- 02The workflow
- 03The data model
- 04The permissions
- 05The API
- 06The assist
A sale closes. Several parties share the commission, a manager must approve it, and VAT applies.
HireLens AIEvidence-based hiring
The AI gathers evidence. A person makes the hire.
The copilot may
- 01Run interviews that adapt to the candidate's CV
- 02Draft evaluation reports that cite their evidence
- 03Answer questions grounded in one job, inside one tenant
Only people decide
Every hire.
Every rejection.
Guardrailed against prompt injection and forbidden decision language.
LinguaCoach AILanguage learning · Arabic speakers
Coaching that keeps teaching when no AI provider is connected.
- A seeded CEFR curriculum — courses are derived, so nothing drifts out of sync
- Realtime coaching streamed over Server-Sent Events
- Queued AI workflows on durable PostgreSQL row leases that survive restarts
- Budgets and daily limits enforced; skipped AI calls are labeled, never faked
A1Beginner
A2Elementary
B1Intermediate
B2Upper-intermediate
C1Advanced
C2Proficient
Inside the product
The learner side and the admin audit trail, on a demo account.
Captured during a live UI audit of a local build, signed in as a seeded demo learner and admin. Activity and accounts are demo data.
Sales intelligence · FastAPI
ProspectIQ
Research → evidence → deterministic scoring → grounded AI → human-approved action.
ProspectIQ
AI can propose. The backend executes.
- Queued
- Running
- Waiting for approval
- Running
- Completed
A consequential action stops the run. After approval the backend reauthorizes the current user, verifies the payload hash and executes once. Then the same AgentRun carries on.
Concept · Real-estate CRM/ERP
ذُرىDhura
An Arabic-first real-estate platform where AI can suggest anything and change nothing on its own.
Concept architecture — not a shipped product
ذُرىConcept architecture — not a shipped product
- 01RTL designed in from the first screen
- 02A deterministic core for listings, deals and commissions
- 03Every AI action waits for a human approval
Applied AI, engineered like the rest of the system.
The same principles show up in every system where I've put a model into a real workflow.
- 01
AI proposes. People decide.
Hire and reject decisions in HireLens are human-only. In Dhura, no AI suggestion changes state without approval. In ProspectIQ an approved action is reauthorized against the current user before it runs.
Seen inHireLens AIDhuraProspectIQ
- 02
Fail closed, not quietly.
ProspectIQ refuses to fall back to demo data in production. CareerGuide won't start if its selected provider has no key.
Seen inProspectIQCareerGuide AI
- 03
A deterministic core.
CareerGuide's scores come from a fixed, version-controlled rules engine. The model advises beside it, never inside it.
Seen inCareerGuide AI
- 04
Honest fallbacks.
LinguaCoach keeps its whole curriculum usable without a paid provider, and labels any AI call it had to skip.
Seen inLinguaCoach AI
- 05
Guardrails live on the server.
In HireLens every request passes through Laravel's provider router and guardrails. The browser never talks to a model.
Seen inHireLens AI
Where the work happened.
Sep 2026 — Present
ArchitWeb
Backend Developer
PHP/MySQL and GraphQL work on operational commerce and POS systems: payment-gateway modernization, refund-error handling, safer wallet-refund retries, and idempotent, indexed reporting APIs.
Jan 2025 — Aug 2026
Rakez Real Estate Company
AI & CRM Specialist / Full-Stack Developer
Laravel/Vue features and backend APIs for a Saudi real-estate ERP/CRM covering leads, sales, contracts, commissions and reporting, with authorization, background processing, integrations and AI-assisted workflows.
2025 — 2026
Business Flow
Full-Stack Developer
NestJS/TypeScript tenant-aware SaaS APIs, role-based business workflows, validation, database structures and integrations with React dashboards.
Jan 2025 — Nov 2025
Mermez
Backend Developer, iLogistics
Laravel/MySQL logistics APIs and workflows for shipment tracking, authentication and reporting, with faster recurring queries through indexing, query refinement and Redis caching.
2022 — Present
Independent
Independent Software Engineer
Custom applications and technical prototypes across SaaS, business operations and AI-enabled workflows, from backend design and APIs through testing and integration.
BSc, Information Engineering & Distributed Systems — Al-Shahbaa University, 2021–2026
More systems.
Smaller chapters, same discipline.
CareerGuide AI
Career matching where the score is deterministic and AI only advises.
Public screens from a local build.
ALGOAG
Write-up in progress
Public screens from a local build.
RestoCafe OS
Multi-branch restaurant POS: offline-first PWA, real-time kitchen display, optimistic concurrency.
Captured by the repo's own Playwright visual tests against a seeded local build. Orders, tables and amounts are test data.
iLogistics
Multi-tenant logistics and booking platform with JWT, role-based access and reporting.
Admin and customer screens from a local build. Names, shipments and amounts are seeded test data.
AgentOS
Details on request
Public pages from a local build, captured with Playwright. The workspace itself sits behind sign-in.
mtjri
Multi-store commerce with a self-service web installer.
The platform admin, a store's back office and its storefront, from a local build with demo stores and orders.
- Business Flow CRMCRM systems and operational dashboards.
Better systems start with understanding the business.
Open to remote roles in backend engineering, business systems and applied AI.