Biography
I am an automation consultant working at the intersection of technology, product development, and process automation. As Product Manager at UNICEPTA, I automated core business processes with agentic methods and took them into production — from large-scale text data mining to client-facing products.
I earned my PhD in Computational Social Sciences at the University of Oxford (with distinction). In my dissertation, I analyzed 40,000 media articles using Python and natural language processing methods; the work was published as a monograph by Palgrave Macmillan in 2022.
My work brings together two strands. For more than a decade I worked as a consultant in foreign and European policy: presenting studies to the Committee on Foreign Affairs of the European Parliament, briefing members of the German and European Parliament, government officials and the German Foreign Office at the German Institute for International and Security Affairs (SWP), and designing international comparative research projects for public and private clients. That work set the standard I still apply — a defensible answer, delivered on time, with its sources named and its limits stated.
As Product Manager I brought that standard to automation. I identified the processes that consumed expert time without rewarding it and rebuilt them end to end, in production:
- AI-based text data mining at scale — automated collection, classification and analysis of large media corpora
- Automated report creation, replacing standardized reporting that had been assembled by hand
- Internal tools and client-facing products built on those pipelines, from prototype to live service
- Prompt engineering and management as a discipline: development and optimization (manually and automated via DSPy), systematic evaluation, documentation and versioning, coordination with teams and clients, and training
- Custom pre- and post-processing inside agentic pipelines — dynamic prompting, context injection, prompt chaining — to make outputs reliable enough to ship
- Proficient across the agentic tooling stack: n8n, CrewAI, LangChain/LangGraph, Microsoft AutoGen, DSPy, Claude Code and Model Context Protocol (MCP) integrations
That combination is what I offer as an automation consultant: I read a domain closely enough to know which judgments must stay with a person, and I build the pipeline that clears away everything around them. I also work with retrieval-augmented generation (RAG) pipelines and have built chatbot prototypes integrating text chunking, the OpenAI Embedding API and Pinecone as a vector database.
