Philipp Dowling

Entrepreneur, NLP Researcher, Creator.


I’m interested in exploring how technology, design, and philosophy can help us better understand the world around us. In my view, many open problems in this domain relate to understanding the structure of language and how we can use computing to make sense of it at scale.

I am a co-founder of, where we work on augmenting human intelligence to empower analysts and allow them to work on more important tasks than gathering data.

I’ve previously spent time doing research in NLP and its intersection with HCI at HKUST, UC Berkeley and Stanford. I completed my MSc in Informatics at the Technical University of Munich in 2018, along with an Honors Degree in Technology Management at the CDTM in Munich.

I also like skateboarding, hiking, hip-hop and rap, playing drums, and obscure memes.

Screenshot 2019-06-20 at 19.29.42.png

Selected Work


I founded together with Hai Nguyen Mau and Markus Müller. Our vision is to make the world’s business information accessible at scale. We’re active in the spaces of Consulting, Risk Management, Alternative Data, ESG, and Corporate Intelligence.

To find out more or to work with us, check out our website (or contact me directly).


I like to work on interdisciplinary projects, especially on applying NLP to domains where users tend to be non-technical. Projects listed here are mostly private work or collaborations, some are from Hackathons, some might be academic or professional work.

EU Discourse Analysis

EU Discourse Analysis





See more projects here.


Master’s Thesis

My M.Sc. thesis “Effective and Scalable Sentence Representation through Dynamic Grassmannian Ellipsoids” was completed in 2018 as joint work between TUM and Stanford, advised by Prof. Stephan Günnemann at TUM, as well as Dr. Pablo Paredes of the Stanford HCI Group. The work deals with a novel approach to representing natural language sentences and evaluating their similarity, efficient neighbour search in this representation space, and introduces an interactive downstream application based on the representation approach.

Download: Thesis / Slides / Code

Screenshot 2019-06-21 at 14.40.05.png

Other work

“Qualitative Exploration and Early-discovery of Prior Knowledge for Causal Inference“ - Paredes et al. Explores an approach to conducting causal inference on textual data collections through semantic sentence embedding methods in a scalable retrieval system. Unpublished (accepted to OSSM 2017 but ultimately not presented). Download

“bAbIxyz - Benchmark tasks for large-vocabulary QA“ - Philipp Dowling and Hinrich Schütze. Introduces a more difficult version of Facebook’s bAbI tasks and evaluates a character-level seq2seq approach baseline on the task. Work done as part of my interdisciplinary project at TUM, together with Prof. Schütze at LMU. 2017. Download. Code.

“Improving evaluation and optimization of MT systems against MEANT“ - Lo et al. Applies vector embeddings to MEANT, a system for the evaluation of machine translation systems developed at HKUST. Proceedings of the Tenth Workshop on Statistical Machine Translation, 2015. Download

“Distributed Vector Space Models for Semantic MT Evaluation in MEANT“. My B.Sc. thesis conducted at HKUST and TUM, advised by Prof. Dekai Wu and Prof. Thomas Runkler. 2015. Download