§ 03 · Selected Work & Research Archive
Projects & Research
Engineering implementations, industry partner collaborations, and academic research focusing on LLM agent systems, foundation model forecasting with TabPFN, Bayesian customer modeling, and graph machine learning.
-
Game revenue forecasting with foundation models
Evaluated whether planning text informs TabPFN-TS net revenue forecasts across 49 monthly origins, comparing LLM extractions, raw-text features, and embeddings against calendar and shuffled-text controls.
-
Agentic toolkit for Power BI report changes
Multi-agent Claude Code toolkit used across the analytics team at InnoGames, planning, implementing, and verifying Power BI dashboard changes in PBIP/TMDL without step-by-step human input.
-
ChefTreff AI Hackathon: Founder Assistant
Multi-agent assistant that helps founders launch a company in Germany, covering automated business plan drafting, marketing strategy, and legal document preparation.
-
Knowledge Graph Question Answering with LLMs
SPARQL-based question answering over structured knowledge graphs, translating natural language questions into executable graph queries over RDF knowledge bases.
-
Lüneburg Public Bus Routing with HVV Timetable Data
Applied machine learning pipeline to model and predict local bus routes for regional transport operator KVG, fusing public transport geospatial data with predictive models.
-
Artist influence networks with Neo4j
Graph database analysis using Neo4j and Cypher to investigate musical lineage, community clustering, and quantitative influence centrality across artists.
-
Hierarchical Bayesian Pareto/NBD customer modelling
Implemented and validated the hierarchical Bayesian Pareto/NBD model from Abe (2009) on the canonical CDNOW dataset using MCMC sampling in PyMC.
-
Broken-tile detection with YOLOv8
Built in 24 hours: a computer vision object detection and classification pipeline for identifying broken, chipped, or damaged roofing materials from product imagery.