Automation engineer · Cairo · Open to work

Manual work is unclaimed profit.

I look at how work actually gets done, find the hours leaking out of it, then build the thing that takes them back.

Scripts, data pipelines, AI agents, internal tools. I've built over a hundred of them in five years of freelancing, while I was studying computer engineering.

inboxspreadsheetsAPIsscrapersagentreclaimed

What reclaiming a day looks like

4 hours a day4 minutes

Someone spends a morning on inbound mail, or rebuilds the same report by hand every week. Once the flow runs itself, that work is a few minutes of checking exceptions.

Figure measured on a client's inbox triage, before and after.

100+

Automations delivered

5

Years freelancing, while studying

8

Languages in production use

1

Mars rover, Kraków

Selected work

Three that show the range.

01

RUNE

An open source platform for building automations. I built the execution core: the small language that reads a node graph, the schema that stores it, and the Go workers that run it all at once.

GoDSLPostgresMicrosoft sponsored
View the repository

Friction

Automating anything still assumes you know distributed systems. The tools that hide that, like Zapier and n8n, put it behind a subscription that gets expensive fast.

Fix

A visual canvas backed by a small language, run by workers that execute nodes concurrently, with each node's state visible while it runs.

Result

Free and self hostable, so the bill doesn't climb at exactly the moment the automation starts being useful.

A RUNE workflow running on the canvas, nodes lighting up as the flow executes
02

Gmail customer data bot

A firm had someone reading every inbound email and retyping customer details into a database by hand.

PythonOpenAIGmail API

Client work, so the code and the screenshots stay private.

Friction

Four hours of somebody's day, every day, and the data still arrived late and inconsistent. The real cost was the salary plus every mistake that followed downstream.

Fix

An agent reads each message, works out what kind it is, pulls the fields that matter, checks them, and writes straight into the database.

Result

Four hours a day became a few minutes of looking over the exceptions. Nobody retypes anything now.

03

ASU ROAR

System architect on a competition Mars rover. Same instinct, pointed at hardware: find the broken link in the chain, then fix the chain.

ROSCANRF meshEuropean Rover Challenge

Friction

Localisation and mapping were dead because the ROS transform tree was malformed, and the command link kept dropping in the middle of a run.

Fix

Rebuilt the transform tree, wrote a UART to CAN interfacer for the Jetson, then designed and RF tested a mesh network for the link.

Result

Mapping came back, the link held through the run, and over a hundred recruits went through two workshops I wrote.

Patterns

What I'm hired to do.

Most of the hundred plus projects are undocumented or under NDA, but the shapes repeat. These four cover nearly all of them.

Instant reporting

Staff assembling the same PDF report by hand, every single week.

Now it generates on demand and mails itself out, and the hours that used to go into it go somewhere useful.

Data pipelines

Copy and paste between an API, a spreadsheet, and a calculation nobody quite trusts.

Collected through an API or a scraper, cleaned up, run through the real calculation, and published as the output people actually asked for.

Spreadsheet to app

A fragile workbook that only one person in the company can safely open.

Same logic, but behind a real interface the whole team can use without breaking anything.

Process audit

“We know it's slow, we just don't know where.”

I map how the work really flows, then tell you honestly which parts are worth automating and which ones aren't.

Method

How the hours come back.

01

Audit

Sit with the process as it really runs. Time it. Find where the hours are going.

02

Map

Draw the flow people actually follow, not the one on the org chart. They usually differ.

03

Automate

Scripts, APIs, agents, workers. Whatever the job needs, and nothing it doesn't.

04

Hand over

Documented, owned by them, still running long after I've moved on.

What I build with

Languages

Python
Go
C and C++
JavaScript
TypeScript
Java
Verilog

Automation and AI

OpenAI API and agents
Google Apps Script
Selenium
Workflow engines

Cloud and data

AWS, Lambda
Docker
PostgreSQL
MongoDB

Hardware and systems

ROS
CAN bus
Tiva C, RISC-V
Linux

About

A computer engineer who kept taking freelance work.

I graduated in computer engineering from Ain Shams this year, and I freelanced the whole way through. That taught me something a degree doesn't: how to build for people who don't care how it works, only that their afternoon is free again.

I like hard, niche problems, and I pick things up quickly. Apps Script, then Python, then agents, then Go workers and a rover's transform tree. I'm a recovering perfectionist, so shipping on time is the thing I actively practise. And I do my best work on a good team.

Abdelrahman Hany

Credentials

BSc Computer Engineering Ain Shams University, 2026 · GPA 3.42
AWS Cloud Architect · NTI, 2025
MIT 6.86x Machine Learning · 2024Verify
MIT 6.431x Probability · 2024Verify
Harvard CS50 · 2022Verify
Cisco CCNA · 2023Verify
2nd place, ASU AI Engineering Hackathon · 2024See it
NASA Space Apps, climate dashboard · 2024See it

Engineering foundations

University and competition work. Not value recovery projects, but where the fundamentals came from.

RISC-V processor in Verilog, 30+ instructions

CHRONOS, a CPU scheduler with live Gantt charts

CartNest, sharded full stack e-commerce

GPS tracking on Tiva C, MCU to server API

NASA Space Apps 2024, a 20 year climate dashboard

Contact

Where is your team losing hours?

Tell me about the process. I'll tell you honestly whether it's worth automating, and what it would take.