Philipp Bongartz
Bio
Updated 03/23/26Philipp Bongartz is a German machine learning engineer and independent AI alignment researcher who received a Long-Term Future Fund grant of approximately $25,000 for prosaic alignment research using chess as a model domain. He has developed a research agenda centered on building a multi-modal chess-language model with an encoder-decoder architecture, using chess as a testbed for investigating symbol grounding and truthfulness in AI models. His GitHub projects include Chess2Vec, ChessTransformer, and other chess-related ML tools, as well as bioinformatics work suggesting a computational background spanning machine learning and life sciences. He holds a PhD (defended in early 2020) from the Heidelberg Institute for Theoretical Studies (HITS). He has been an active contributor to the LessWrong and Alignment Forum communities under the handle p.b. since December 2020, with over 26 posts and 295 comments on topics ranging from chess AI to model scaling and alignment methodology. He has worked as a Senior Consultant and Data Scientist at Exxeta AG in Germany and is a rated chess player with a FIDE standard rating of approximately 2155.
Community Signal
Updated 03/23/26No endorsements yet.