Understand your data landscape with Observability platform
Designing data observability in Azure so teams know what data exists, how it moves and whether it can be trusted.
About
Maciej Kępa - Data & AI Architect
I’m Maciej Kępa, a Data & AI Architect at Datumo, based in Kraków, Poland. I work on MLOps and data platforms, primarily with Azure and Databricks, focusing on how teams deploy, monitor and maintain data and ML systems.
My work centers on making AI useful in real environments: platform architecture, repeatable delivery, MLOps foundations, observability and the engineering boundaries between data systems and machine learning systems.
Azure-native data platforms and delivery systemsDatabricks, lakehouse patterns and analytics-ready architectureMLOps foundations for repeatable delivery and model operationsData engineering connected to production ML and AI systems
I care about systems that are scalable, observable, maintainable and actually ready for production, not only persuasive in a workshop or notebook demo.
Speaking and community
Sharing the work
I am a member of Data Community Poland and part of the SQLDay organizing team. My talks and workshops connect architecture decisions to operating data and ML systems.
MLOps on Azure - war stories and lessons learned
Operationalizing ML in regulated and edge-heavy environments through real production lessons from Azure-based systems.
Trenuj, wdrażaj, nadzoruj: techniczny przewodnik po AI w Databricks
A full-day workshop following data preparation, model training, pipelines and model operations in Databricks, with hands-on exercises.
What I build
Three recurring problem spaces
The common thread is operational reality: architecture, delivery and platform quality need to hold together under production constraints.
Data Platforms
Designing Azure- and Databricks-based platforms that make analytics and AI delivery operational, not improvised.
MLOps Systems
Building model lifecycle foundations, delivery workflows, observability and reliability around production ML.
AI in Production
Treating model quality and platform quality as the same engineering problem, especially in edge, IoT and data-centric workloads.
Principles
How I tend to reason about systems
01
Prefer systems that can be operated by teams, not only by their creators.
02
Bias toward reproducibility, observability and explicit trade-offs.
03
Treat notebooks as a useful beginning, not a deployment strategy.
Beyond work
Interests that map well to engineering
Animal lover with a soft spot for real-world chaos over perfect lab conditions.
Gaming enthusiast, especially where systems, strategy and immersion matter.
DIY hobbyist who likes building, fixing and understanding things end to end.