I’m Abdullah Abdelrazek, based in Germany. I build AI products with large language models, design the cloud they run on, keep it secure and lead the team that ships it.
When a project grabs me, I spend nights, holidays, whatever it takes to get it done.
I love learning new things. Most of my projects start that way.
What result do you want, and how will we know we got it? We write it down together before anything gets built.
Took KLAO from a first prototype to paying public-sector clients, with an EXIST grant and Media Lab Bayern behind it.
Your users, your process, your data and your rules. I listen first, then design a workflow your team can run.
Technical counterpart for clients’ engineering teams while they integrated the KLAO API into their own systems.
A rough version on your real data within days. Nobody knows the quality an AI can reach until we test it.
KLAO started as a rough prototype on real public-sector texts, in front of its first users before much else was built.
What we build, what is out of scope, the risks, and a test set we agree on. Then a green light.
KLAO set its bar early: which texts, which Leichte Sprache rules each must follow, and a human review before certifying.
Fine-tuning, retrieval and the backend around them, built in short loops with regular demos along the way.
Master’s thesis on Leichte Sprache, rated well above the compared models, and multimodal retrieval for a bearing manufacturer.
Cloud, security and data protection are built in from the first day, not patched on at the end before launch.
KLAO scales on AWS and is GDPR compliant by design. I secured it with lessons from finding bugs at Uber, IBM and HackerOne.
User feedback becomes test cases, and every drop in quality gets traced back to its cause and fixed.
KLAO’s MVP went out to its first users early, and their feedback, not guesses, decided what got fixed first.
Your team gets the code, the know-how and the docs, so it can run and grow the product without me.
In use at the Stadtschulamt Frankfurt, the Landratsamt Starnberg, the University of Würzburg and municipal museums.
Media Lab Bayern Media Startup Fellowship
Then the Scale Up programme
Jury Award
Special prize for social entrepreneurship
Honourable mention
Happy to talk about either of these roles.
For an AI product that needs someone to take it from the first idea to the launch, the whole A to H.
Tell me about your productWorking inside a client’s team: learning their workflow, building on their data and handing over something they can run.
Tell me about your team