// data & ai platform architect
Roland Utz
I architect and build enterprise data platforms and production AI systems — from streaming pipelines to LLM applications — for banking, telecom, energy, and industrial clients.
what i do
Architecture, hands-on engineering,
and technical leadership.
Physicist by training, engineer by trade. Fifteen-plus years as an independent contractor delivering large-scale data and machine-learning systems to enterprise standards of governance, compliance, and scale.
01 / data-platform-engineering
Data Platform Engineering
End-to-end pipelines and cloud big-data platforms at enterprise scale — streaming, warehousing, and governance built to survive production.
Spark · Kafka · Flink · Snowflake · Airflow
02 / generative-ai-llm-systems
Generative AI & LLM Systems
Production NLP, deep learning, and LLM applications — RAG platforms, model selection, fine-tuning, and evaluation strategies that ground answers in facts.
PyTorch · Transformers · vLLM · Milvus · Qdrant
03 / distributed-systems-cloud
Distributed Systems & Cloud
Cloud-native architectures across Azure, AWS, and GCP — Kubernetes, high-throughput messaging, and migrations off legacy on-prem ecosystems.
Azure · AWS · GCP · Kubernetes · Docker
selected work
Systems that shipped.
Enterprise Software Vendor
Enterprise RAG Platform
Greenfield Retrieval-Augmented Generation platform, from concept through MVP: ingestion, embeddings, vector search, and LLM orchestration across enterprise knowledge sources.
Python · PyTorch · vLLM
Major German Bank
Big Data Platform & Pipelines
Distributed Spark pipelines for the bank’s next-generation big data platform, with the technical evaluation that established Spark as its core compute engine.
Spark · Scala · Hadoop
European Telecom Group
Cloud Big Data Migration
Migration of a legacy on-premises Cloudera ecosystem to a cloud-native Azure architecture for high-volume telecom streams, including a custom Spark geospatial library.
Azure · Kafka · Kubernetes
Industrial SMB
IIoT Anomaly Detection
Sensor analytics and anomaly detection for industrial IoT, designed, productionized, and cost-optimized on AWS.
Python · TensorFlow · AWS
from the blog
Notes from the field.
August 25, 2026
When Your RAG Knows It's Wrong
Part 3 of the RAG Architecture Series. Corrective RAG and Self-RAG are two different answers to the same question: can the pipeline notice its own failure? One changes the pipeline. The other changes the model. Here's how they behave once real traffic hits them.
read
August 23, 2026
Late Interaction, SPLADE, and the Architectures I Still Do Not Ship
Addendum to Part 2. The four families between a bi-encoder and a cross encoder, what each one would have to prove, and the number that keeps deciding it for me.
read
August 16, 2026
Hybrid Retrieval and Reranking: The Part Where We Actually Build It
Part 2 of the RAG Architecture Series. Chunking, BM25 analyzers, fusion weights, cross encoders, and the mechanism behind the part number example I left dangling last time.
read