Paul Roeseler (email)

Currently the CTO of Swisscoding Technologies in Zurich.

My Experience

• CTO at Swisscoding Technologies building clinical-coding LLM agents (RL post-training, production inference, synthetic data).
• Senior AI Engineer at IBM working on multimodal document AI + agentic RAG and LLM post-training (distillation / DPO).

My Projects

• Tournament Reranker - dependency-light Python library for LLM reranking using a tournament / pyramid strategy. (Python, LLMs)
• ORFA / IORFA — An Optimal RuleFit Algorithm - optimal-tree RuleFit + single-step mixed-integer formulation. (Python, MIO, interpretability)
• helpit - LLM-powered wrapper around Python’s help() that lets you ask questions about concrete objects with doc-grounded answers.

Misc

• Mostly: LLM post-training, optimization, document intelligence, infra for training/inference.