About Daniel Magyar
Most AI projects die between the demo and production.
Not because the model doesn’t work, since the demo already proved it does. They die on the unglamorous parts: integrating with systems that already exist, keeping costs predictable, making failures visible at 3am, and answering where the data is legally allowed to live.
That’s the work I do.
What I build
I’m a cloud and GenAI engineer at ApexLab, an AWS consultancy. Most of my work is generative AI on AWS: Bedrock, Lambda, DynamoDB, serverless architectures built to survive real traffic and real auditors.
A recent example: a document classification system for a healthcare client, processing sensitive material under constraints that ruled out most of the obvious approaches. The model was the easy part.
How I work
I build infrastructure I can recreate from scratch. Terraform and Pulumi for environments, Kubernetes only where it earns its keep, CI/CD that people trust enough to use without thinking about it.
AI systems are still systems. They need monitoring, cost controls, rollback paths, and someone who knows what happens when they fail.
I wrote up doing exactly that for this site itself: rebuilding danielmagyar.com.
What I’m interested in
Where AI workloads are allowed to run.
Data residency, the European sovereign cloud landscape, and the regulatory pressure (NIS2, DORA, CRA) that increasingly decides architecture before anyone writes code. Most writing on this is aimed at compliance teams. I write it for engineers, because by the time it reaches us it’s an architecture problem, not a policy one.
Currently
- Cloud & GenAI engineer at ApexLab
- AWS Certified Solutions Architect – Associate
- Working toward AWS Certified Generative AI Developer – Professional
- Based in Budapest
Away from the keyboard
Hiking, and cooking things I haven’t cooked before. Both involve following a plan until it stops working and then improvising, which is roughly the job description.
Work with me
I take on consulting work in GenAI architecture on AWS, cloud infrastructure, and data residency strategy.