Ioannis Papadimitriou

I’m a Machine Learning Scientist working at the interface of machine learning and molecular/physical systems, where I integrate physical principles with data to deliver reliable, high-performing models. I design and ship deep-learning systems across molecular graphs, spectra/images, and text, including physics-informed and graph neural network approaches, and I adapt foundation models for demanding scientific problems.

I build ML-accelerated surrogates for expensive atomistic and quantum-chemistry simulation, owning projects end to end: framing the question, implementing in PyTorch, defining benchmarks and evaluation, and delivering production-grade, reproducible pipelines. At CERTH, I lead deep-learning research that connects problem formulation through benchmark definition and validation, including real-time inference and rigorous RAG evaluation frameworks.

My work has resulted in 27 peer-reviewed papers, 1,100+ citations (h-index 19), and authorship of the field’s most-cited recent review on AI for materials discovery (161 citations). I also help teams and EU research consortia move from research prototypes to systems that scientists can trust and deploy.