Tim Salimans
Member Of Technical Staff at Anthropic
Randstad, Netherlands
Summary
🤩
Rockstar🎓
Top SchoolTim Salimans is a machine learning research scientist and team leader with 12+ years of experience building and improving generative models, variational inference, and practical deep learning systems. He helped pioneer practical VAE reparameterization techniques (earning the Lindley Prize) and is widely known for influential GAN work, including semi-supervised applications and the Inception score, with code contributions to high-profile repos like OpenAI's improved-gan. Tim has led research teams at DeepMind and OpenAI, founded AI startups that outperformed human experts in medical imaging, and continues to tackle engineering-scale problems such as memory-efficient training (gradient checkpointing) for very large neural networks. Equally comfortable in theory and production, he combines a PhD in econometrics with hands-on success in Kaggle competitions and building industry-grade ML tooling.
12 years of coding experience
14 years of employment as a software developer
Exchange Semester in Australia, Science, Exchange Semester in Australia, Science at Monash UniversityPhD, Econometrics, PhD, Econometrics at Erasmus University RotterdamBSc (Hons), Liberal Arts and Sciences (Magna Cum Laude) Major in Mathematics & Physics, BSc (Hons), Liberal Arts and Sciences (Magna Cum Laude) Major in Mathematics & Physics at University College Utrecht
Github Skills (19)
scipy10
python10
dcgan10
machine-learning10
cgan10
lasagne10
tensorflow10
paper10
neural-network10
theano10
numpy9
maths9
mathics9
math9
mathematics9
Programming languages (4)
Github contributions (5)
openai/improved-gan
Jun 2016 - May 2018
Code for the paper "Improved Techniques for Training GANs"
Role in this project:
ML Engineer
ML EngineerContributions:23 commits, 2 PRs, 5 pushes in 1 year 11 months
Contributions summary:Tim contributed to the `openai/improved-gan` repository, which focuses on improving GAN training techniques. Their commits primarily involve modifications to the `nn.py` file, suggesting they worked on core neural network components. These changes included adding, removing, and refactoring various network layers and functions. Furthermore, the user made updates to training scripts, indicating involvement in the model training process.
pytorchtechniquesimproveddeep-learninggenerative-adversarial-network
cybertronai/gradient-checkpointing
Oct 2017 - Jan 2018
Make huge neural nets fit in memory
Role in this project:ML Engineer
ML EngineerContributions:28 commits, 3 PRs, 10 pushes in 3 months
Contributions summary:Tim primarily focuses on improving the memory efficiency of gradient computation within a TensorFlow environment. Their contributions involve modifying existing code and implementing techniques for gradient checkpointing. This includes the implementation of automatic checkpoint selection strategies, modifications to the gradient computation process, and adjustments to improve test coverage and correctness. The user's work directly impacts the ability to train large neural networks within memory constraints.
memorydeep-learningnetsneural-networksfit
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.