Arthur Hemmer, PhD - Applied Scientist & Neurosymbolic AI Researcher

Applied Scientist based in Paris specializing in Neurosymbolic AI and reliable, production-grade systems.

Background

Over seven years of experience bridging theoretical research and industrial application. Currently at Shift Technology, leading technical initiatives to automate high-stakes decision-making in the insurance sector.

Projects

TimestampAI - Video navigation tool
lazyk - C++/Python library for constrained decoding
transactional-documents - Neurosymbolic Information Extraction framework

Publications

Neurosymbolic Information Extraction from Transactional Documents - IJDAR, 2025
Improved Information Extraction by Leveraging Multi-Hypothesis OCR - ICCVW, 2025
Confidence-Aware Document OCR Error Detection - DAS, 2024
Lazy-k Decoding: Constrained Decoding for Information Extraction - EMNLP, 2023

Writing

On the Duality of Intelligence - Reflections on Neuro-Symbolic AI