Anders Andreassen
Senior Staff Research Scientist at Google DeepMind
New York City Metropolitan Area United States
Summary
🤩
Rockstar🎓
Top SchoolAnders Andreassen is a Senior Staff Research Scientist at Google DeepMind with seven years of research experience focused on enhancing reasoning in large language models like Gemini and Minerva. Trained as a theoretical physicist (PhD, Harvard) and seasoned by postdoctoral work at Berkeley, he brings rigorous mathematical thinking to practical LLM training-data and evaluation challenges. He has hands-on experience improving benchmark tooling and model scoring—contributing reliability and batching optimizations to the high-profile BIG-bench project. Based in the New York City area, Anders blends deep theory with production-minded engineering to push model understanding and reasoning forward, and he’s known for improving subtle infrastructure issues that materially raise experimental fidelity.
7 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy (PhD) Physics, Doctor of Philosophy (PhD) Physics at Harvard UniversityNorwegian University of Science and Technology
German, Norwegian, Norwegian, English
Github Skills (8)
huggingface-transformers10
python10
apachebench10
machine-learning9
json9
api-design8
testing6
tensorflow4
Programming languages (4)
Github contributions (5)
google/BIG-bench
Sep 2021 - May 2022
Beyond the Imitation Game collaborative benchmark for measuring and extrapolating the capabilities of language models
Role in this project:
Back-end Developer
Back-end DeveloperContributions:9 reviews, 12 commits, 18 PRs in 8 months
Contributions summary:Anders contributed to bug fixes within the `big-bench` repository, specifically addressing issues related to random number generation in the `JsonTask` class, indicating a focus on improving the task's reliability. The user also made changes to the `huggingface_models.py` file, optimizing the scoring process for Hugging Face models through batching techniques. Furthermore, the user introduced utilities for task name access, enhancing the project's organization and ease of use.
bertmachine-learningbenchmarkmeasuringbenchmarks
andersjohanandreassen/BIG-bench
Jul 2021 - Aug 2022
Beyond the Imitation Game collaborative benchmark for enormous language models
Contributions:2 PRs, 59 pushes, 11 branches in 1 year 1 month
benchmarkcollaborativelanguage-modelsbeyond
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.