Oyvind Tafjord
Staff Research Scientist at Google DeepMind
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolOyvind Tafjord is a staff research scientist at Google DeepMind with 12+ years bridging research and engineering in machine learning and NLP, following a principal research scientist role at AI2. He holds a PhD in Physics from Princeton and built a career applying rigorous scientific methods to large-scale ML systems, model training, and evaluation. Oyvind has deep hands-on experience integrating C++ toolkits with higher-level languages—evident from his Java bindings work on the DyNet neural network toolkit—and contributed notable features to AllenNLP such as TensorBoard parameter/gradient tracking and improved tokenization. His background spans both product-grade architecture (as Chief Architect at Wolfram Alpha) and core research, giving him a rare fluency across prototyping, productionization, and research publication. Based in Seattle, he combines academic rigor with practical engineering, often focusing on tooling that makes research code robust and observable. A less obvious strength is his long-term habit of turning research-grade libraries into usable developer bindings and monitoring features that accelerate adoption.
12 years of coding experience
26 years of employment as a software developer
Norwegian University of Science and TechnologyDoctor of Philosophy (PhD) Physics, Doctor of Philosophy (PhD) Physics at Princeton University
Norwegian
Github Skills (23)
pytorch10
c-language10
python10
java10
binding10
databinding10
javas10
deep-learning10
nlp10
cprogramming-language10
keybindings10
swig10
bind10
data-science9
model-driven9
Programming languages (6)
Github contributions (5)
allenai/allennlp
Aug 2017 - Jan 2019
An open-source NLP research library, built on PyTorch.
Role in this project:
ML Engineer
ML EngineerContributions:17 commits, 33 PRs, 14 pushes in 1 year 4 months
Contributions summary:Oyvind primarily contributed to the AllenNLP library by implementing features related to model training and evaluation. They added functionality to track parameters and gradients using TensorBoard, enhancing the model's monitoring capabilities. Furthermore, the user made modifications to the code by fixing bugs and improving overall code quality. They also introduced a new word splitter, improving tokenization capabilities and refactored several parts of the library.
nlppytorchtransformerspythonbert
clab/dynet
Nov 2016 - Jan 2017
DyNet: The Dynamic Neural Network Toolkit
Role in this project:Back-end Developer
Back-end DeveloperContributions:17 commits in 2 months
Contributions summary:Oyvind contributed to the Java bindings of the DyNet deep learning library. Their work involved creating a proof-of-concept Java binding, which included defining necessary headers, wrapping C++ code with SWIG, and creating examples. The commits demonstrate the integration of DyNet's core functionalities like model creation, computation graphs, and training within a Java environment using SWIG. This involved extending the binding capabilities to support new functionalities, alongside fixing and enhancing the existing integration.
dynetdynamic-neural-networkdeep-learningneural-networksmachine-learning
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.