ML/AI Projects

This page collects applied work, open-source projects, and experiments from my work with machine learning, LLMs, and AI-assisted work. Some originated in professional settings; others began with my own questions or as learning experiments.

Selected projects

Local LLM Benchmark Suite A public benchmark suite for my actual writing, knowledge, coding, and agentic tasks. The goal is to derive useful routing rules instead of naming one supposedly best model.
LLM-based processing of tender documents Automated support for the sales and quotation process at a market-leading midsized mechanical engineering company.
Predicting touch time in software development An experimental ML approach with TechDivision based on issue-complexity categories.
RAG experiment with LangChain, Llama 3, and Chroma A publicly reproducible experiment with retrieval-augmented generation and local AI.

More projects and experiments

Tooling for my former AI newsletter An early local LLM pipeline for RSS selection, web scraping, and summarization. Archived project.
Serverless Gemini Chatbot Client-side JavaScript with a Google Cloud Run function acting as a proxy to Vertex AI. Archived experiment; source code available, no live demo.
Predicting the energy output of a combined-cycle power plant A graded project in the Duke University course “Machine Learning Foundations for Product Managers.”
Regression experiment with UCI data A comparison of an artificial neural network and a random forest.
Testing the diversity prediction theorem An experiment using different ML algorithms.

Workshops and community

AI Transformation Camp An unconference for people who want to make a practical impact with AI in organizations.
AI literacy and competence under Article 4 of the EU AI Act My three-hour workshop on using AI and GenAI at work—including risks and legal and societal implications.
ChatGPT & Co. for Product Owners and Product Managers My eight-hour workshop for product professionals on using GenAI in their daily work.