Jason Ramapuram
Research Scientist at Google DeepMind
United States
Summary
👤
Senior🎓
Top SchoolJason Ramapuram is a research scientist with 11 years of experience building and shipping machine learning systems across industry and academia, currently at Google DeepMind after six years as an ML research scientist at Apple. He holds a PhD in Machine Learning from the University of Geneva, where he produced work on differentiable latent memory and scalable variational inference, and has publications at ICLR, NeurIPS and ECML PKDD. Jason combines deep algorithmic expertise (autoencoders, variational methods, associative/holographic memory, LSTMs) with low-level engineering in C/C++/CUDA and a growing Rust toolchain, authoring a cross-GPU Rust ML library and contributing substantial features to Keras and arrayfire-rust. His background spans real-time streaming systems, embedded and production deployments (Qualcomm, Viasat) and open-source contributions, reflecting a rare blend of research rigor and production-grade software craftsmanship. Notably, he has extended core deep-learning tooling (Keras autoencoders) and worked on self-supervised multimodal models that reached NeurIPS, showing impact from prototype to community-facing projects.
11 years of coding experience
16 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Machine Learning, Doctor of Philosophy (Ph.D.), Machine Learning at University of GenevaMasters of Science, Electrical Engineering, Masters of Science, Electrical Engineering at University of California, Riverside
Kannada, English
Github Skills (9)
neural-network10
keras10
machine-learning10
deep-learning10
tensorflow10
python10
autoencoder10
pytorch9
data-science8
Programming languages (13)
Github contributions (5)
keras-team/keras
May 2015 - Jun 2015
Deep Learning for humans
Role in this project:
ML Engineer
ML EngineerContributions:8 commits, 6 PRs, 50 comments in 22 days
Contributions summary:Jason's contributions primarily focused on enhancing the autoencoder functionality within the Keras library. They implemented and tested autoencoder models, including classical and denoising variants, and enabled the use of any type of layer as input. Additionally, the user extended the library to support deep architectures and ensured correct weight tying. These changes demonstrate expertise in deep learning architectures and the Keras framework.
pythondata-sciencedeep-learningneural-networksmachine-learning
jramapuram/SimCLR
Apr 2020 - Jun 2020
SimCLR pytorch implementation using DistributedDataParallel.
Contributions:38 commits, 1 PR, 32 pushes in 1 month
pytorchsimclrpytorch-implementation
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.