01 / 04
Night film
Five shots, one night, no crew. Rain over an old stone city, a boy, and a cage he climbed to the roof to open.
Directed, generated and cut end to end
AI film · 9:16 · 8 s
AI automation engineer & creative technologist
Amman · Remote or relocation
Five systems.
One person.
Lana Al-ZoubiAI automation engineer · Amman
Everything on this page was designed, built and shipped by me. Yours would be too.
Open to full-time roles and long engagements
Start00 / 05
Selected work · 05
03 · Own product · 2026
Rasel
One AI agent across WhatsApp, Instagram and Facebook. It replies in Arabic, logs every lead and follows up on its own.
0Workflow steps
3Channels, one inbox
0hAuto follow-up
Problem
Businesses here sell across WhatsApp, Instagram and Facebook. The messages sit in three separate apps, and anything arriving after hours goes unanswered.
My role
Designed and built the system: the three channels in one flow, the Arabic prompts and the handover package.
Stack
n8n · WhatsApp Cloud API · Meta Graph API · Claude API · Google Sheets
Result
Messages from all three channels are answered in seconds and logged as one lead record. First client pilot next.
How it works
A message arrives on WhatsApp, Instagram or FacebookIntent is classifiedClaude drafts the reply in Arabic, in the shop's toneReply is sent in secondsLead is logged to Google SheetsComplaint? The owner is alertedNo reply after 20 hours? Follow-up with a booking link
Watch it run
Demo recreation with sample data · 43 s · sound on
AI interviewer
01 · Retail · Electronics megastore
ERP ⇄
Store
A supplier portal, and a two-way link between the company ERP and its online store.
0Suppliers on record
0Independent systems
2sNumber validation
Problem
Around 2,000 suppliers submitted parts by hand, each with dozens of specification fields and several reference numbers. Slow, error-prone, and wrong data reached the store.
My role
Designed and built the portal and the integration, and audited the live workflow.
Stack
Next.js · Prisma · n8n · Dynamics 365 F&O · Shopify
Result
A product is entered once, checked before it is accepted, and published only after a person approves it.
What I built
- Four independent automation systems working together: supplier sync, number validation, content generation and publishing. A fault in one does not stop the rest.
- A daily 07:00 sync that pulls supplier records from the ERP. The ERP stays the single source of truth: the portal reads from it and never writes back.
- An account created automatically for every supplier, activated by a registration link valid for 48 hours and a code valid for 10 minutes that locks after five failed attempts. The supplier sets their own password.
- Access revoked automatically when a contract period ends, with no manual follow-up.
- SKU and model numbers checked against the ERP within two seconds of entry, with duplicates refused.
- Cascading, data-driven fields: nothing is typed freely, and each selection narrows the next to what that supplier actually carries.
- Generated English marketing copy and an Arabic version written as native text, not a literal translation, with the supplier's original input kept for regeneration.
- Mandatory human confirmation before publishing. Any regeneration invalidates the previous confirmation, and the product publishes as a draft first.
- A full activity log: every sign-in, submission, edit, approval and rejection recorded with user, time and IP, on individual staff accounts.
- An audit of the live workflow that found and fixed exposed credentials, a test limit left on, a faulty date filter and data that never returned to the ERP.
1 week → 2 min
02 · In-house · Manufacturer, Jordan
1 week
2 min
The month-end accounting close, automated end to end.
2 minClose run time
0Data sources
12Closes a year
Problem
The month-end close took about a week of manual work every month, across cheques, supplier invoices, attendance data, bank withdrawals and manual entries.
My role
Built the close around the company’s own accounting structure, so cheques, supplier invoices, fingerprint attendance data, bank withdrawals and manual entries all feed into one run.
Result
The close now runs in about two minutes.
Also built
A system that turns the founder's voice notes into formal, ready-to-send emails.
What feeds it
ChequesSupplier invoicesFingerprint attendance dataBank withdrawalsManual entriesChecked against the company’s own accounting rulesMonth closed
Rasel
04 · Own product · Live demo
AI interviewer
A live talking avatar that interviews and screens job candidates in Arabic.
24/7Always available
ARInterview language
1stRound covered
Problem
First-round interviews take hours of a hiring team's time.
My role
Built the interview flow and the live avatar setup.
Stack
Tavus · Netlify Functions · JavaScript
Result
A working demo: the candidate joins on camera and the agent runs the interview.
Film & motion
05 · Film · Motion · Brand
Film & motion
AI film, brand films and motion. Directed, produced and cut end to end.
0Films
0Shoot days
9:16Format
02 / 04
Data film
A thermal visual language built for one idea: work done by hand burns red, automated work runs cold. The week falls from 10,080 minutes to 2.
Directed, generated and cut end to end
Motion graphics · 9:16 · 10 s
03 / 04
Product film
A locked camera and a bottle that never moves. Everything in frame is cloth, fluid and rigid body simulation, cut as a seamless loop.
Directed, generated and cut end to end
3D · Cycles · 9:16 · 10 s
04 / 04
Arabic presenter
A fully generated presenter speaking Levantine Arabic, directed frame by frame to read as a person talking, not a model performing.
Directed, generated and cut end to end
AI video · 9:16 · 10 s
Let's talk
03 · Rasel · Live demo
Talk to the agent
A working copy of the agent, running in your browser on sample data. Pick a channel, send a message in Arabic or English, and watch the lead land in the sheet.
Inbox
Leads · Google Sheets
| Time | Channel | Intent | Status |
|---|
You just ran it on sample data.
Run it on ours01 · Supplier portal · Live demo
Submit a product
Two sides of the same portal: what a supplier fills in, and what the company reviews before anything reaches the store.Short demo · the live portal has many more fields, checks and an audit trail
Supplier side · one of ~2,000 suppliers
This is the screen a supplier gets. Every field is a list drawn from the company ERP, so nothing is typed freely, and the SKU is checked against the ERP before it is accepted.
01 Select02 Validate03 Copy04 Approve
Company side · review before publishing
The company sees the generated English and Arabic copy. Nothing reaches the store until a person confirms both texts, and every action is written to the log below.
Waiting for a validated product.
بانتظار منتج تم التحقق منه.
Activity log
| Time | User | Action | IP |
|---|
You just ran it on sample data.
Run it on ours02 · Month-end close · Live demo
Run the close
At the end of every month the accounts have to be closed: cheques, supplier invoices, attendance, bank withdrawals and manual entries all get matched against each other and against the company rules. By hand it took about a week. Press run and watch the same sequence the real system follows.Short demo · sample month, invented figures, real steps and order
What goes in · the five sources
Each line is a pile of paper or a file someone used to go through by hand.
What comes out
The run time, and the checks the system performs on its way to closing the month.
--:--
Manual close, before: about one week.
| Check | Outcome |
|---|
You just ran it on sample data.
Run it on ours
About
Lana
Al-Zoubi
AI and automation engineer in Amman, working professionally since 2020. I build n8n workflows, AI agents and the web apps and servers they run on.
My edge is the creative side: design, motion graphics, realistic AI video and 3D. It means the systems I build are also clear to the people who use them.
Now
Creative Director & Head of Partnerships, Smart Solidarity. 3D & Multimedia Designer, Yemeni Honey House Factory.
Tools
n8n, Python, Claude API, WhatsApp Cloud API, Next.js, Docker, nginx, Dynamics 365, Shopify.
Study & languages
B.Sc. AI & Data Science, Hashemite University (in progress). Arabic, English.