Skip to content

dataflocks

Data-centered engineering.

I'm Maximilian Karasz, a senior data & ML engineer with 25 years' experience — the last 8+ as an independent practice. I help organisations turn their data, especially the messy, unstructured, under-used kind, into systems they can query and trust. I've done this across pharma, chemicals, insurtech, adtech and energy: graph data modelling, production ML, and technical advisory from the engine room to the C-level.

Read the case study

What I do

  • Graphs & connected data

    Neo4j modelling and graph-based ML, end to end — from refactoring a DAX pharma group's biomedical knowledge graph to graph embeddings for drug-target discovery and value-chain analysis in chemicals. Certified Neo4j Professional with a Graph Data Science certification.

  • Production ML & MLOps

    Data ingestion, retraining and serving ML models in high-throughput services, with CI/CD and automated testing. At an adtech subsidiary of a multinational telecom I cut the training hardware footprint by 75%.

  • Cloud or local — your call

    I've shipped plenty of both. Cloud when scale and speed win; on your own hardware when privacy, sovereignty or GDPR demand it — no data leaves the building. The data and the requirements decide, not dogma. Either way, every answer traces back to its source.

  • Advisory, engine room to boardroom

    From self-service ingestion that lets non-technical colleagues add data sources themselves, to data-lakehouse POCs and technology decisions briefed to data leads and C-level. I pick the right tool for each job — and keep cost honest.

  • Podcasts as Data

    Turning a German economics & politics podcast — hundreds of hours of audio — into a private, queryable knowledge graph that runs entirely on local models. Ask it how a debate shifted over the years and get answers with citations back to who said what, when.

    Read the case study

dataflocks is an independent engineering practice. If you're sitting on data you wish you could query — or wondering whether an AI system can be made trustworthy enough to act on — that's the kind of problem I like.

How this was made. The pipeline and the articles you're about to read were both created with AI assistance. Responsibility for both, and especially for the statements made in the articles, lies with me. I acknowledge and respect that many people, for ethical or other reasons, will not want to consume content that was created in part or wholly using LLMs. This is your notice. My intent is to inform, not deceive.

Close