Janusch Koza

Development of a microservice-based processing system for content from social networks (X/Twitter) and news agencies (dpa-compact, dpa-profeed).

Core tasks: AI-powered classification of incoming content via AWS Bedrock, transformation into structured Markdown format, design of a RAG pipeline for semantic document processing, integration of SQS and Firebase for subscription mechanisms.

Result: Automated processing pipeline with stable, well-tested microservice components as the foundation for a scalable document and news platform in the FinTech environment.

Structured analysis of error-prone import processes in .NET systems – CSV, Excel, XML, JSON. Within 2-3 days, I examine parsing, validation, error handling, and workflow logic, identify concrete bottlenecks, and deliver a prioritized optimization plan.

The result is a clear document: What is the problem, what does it cost, how will it be fixed. Implementation follows upon request.

Typical use: Systems with changing formats, large data volumes, or unstable ETL pipelines.

Targeted integration of AI models into existing .NET applications – OpenAI, Azure OpenAI, AWS Bedrock, Claude, Gemini, or local LLMs.

Process: Analysis of suitable use cases in the existing system, design of an integrable architecture concept, delivery of a working PoC. Where appropriate, workflow automations are integrated via n8n.

Typical use cases: Text classification, document processing, intelligent input assistance, RAG pipelines, automation of editorial or administrative processes.