Sebastian Gehrmann
Head Of Responsible AI, CTO Office at Bloomberg
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolSebastian Gehrmann is a product-minded AI leader with 11 years of experience driving NLP research and responsible AI strategy, currently heading Responsible AI in Bloomberg’s CTO office after leading NLP there. He combines deep academic roots—a Ph.D. from Harvard focused on human–AI text collaboration and tool-building like LSTMVis and exBERT—with industry contributions at Google on model evaluation and PaLM, and community building through the GEM evaluation workshop. Sebastian’s work spans from hands-on model interpretability and attention/copy-mechanism engineering in open-source projects like OpenNMT-py to shaping company-wide policies that make AI systems more trustworthy and transparent. Based in New York, he blends research credibility with product delivery, and he has a track record of turning visualization and evaluation tools into practical resources used by researchers and engineers.
11 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Computer Science, Doctor of Philosophy (Ph.D.) Computer Science at Harvard UniversityBachelor of Science (BSc) Wirtschaftsinformatik (Industrial Engineering), Bachelor of Science (BSc) Wirtschaftsinformatik (Industrial Engineering) at Fachhochschule Nordakademie Elmshorn
Github Skills (14)
attention-mechanism10
pytorch10
machine-translation10
machine-learning10
lstm10
nlp10
recurrent-neural-networks10
deep-learning10
neural-machine-translation10
python10
hdf9
gru8
data-analysis8
tensorflow7
Programming languages (5)
Github contributions (5)
OpenNMT/OpenNMT-py
Jul 2017 - Nov 2018
Open Source Neural Machine Translation and (Large) Language Models in PyTorch
Role in this project:
Back-end Developer
Back-end DeveloperContributions:156 commits, 42 PRs, 9 pushes in 1 year 4 months
Contributions summary:Sebastian's primary contribution involved modifying and enhancing the implementation of the attention mechanism, specifically focusing on integrating the Bahdanau attention type within the `GlobalAttention.py` module. This included refactoring the code to align with PEP8 standards and adapting option names for the attention type in `train.py`. Furthermore, the user made several additions and improvements to the copy mechanism which would support the specific needs of the OpenNMT-py project.
pytorchdeep-learningmachine-translationtranslationneural-machine-translation
HendrikStrobelt/LSTMVis
Jun 2016 - Apr 2018
Visualization Toolbox for Long Short Term Memory networks (LSTMs)
Role in this project:Data Scientist
Data ScientistContributions:1 release, 36 commits, 1 PR in 1 year 11 months
Contributions summary:Sebastian's primary contribution involves implementing and refining machine learning models, particularly focusing on Long Short Term Memory (LSTM) networks, as suggested by the repository description and the presence of LSTM-related scripts and configurations. They developed scripts to extract model weights, analyze hidden states, and compute saliency/influence scores for words, demonstrating a strong focus on model interpretability. The user further optimized existing scripts for memory efficiency and added support for GRU models.
termmemorydeep-learninglstmrecurrent-neural-networks
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.