Neel Kovelamudi
Research Engineer at Google DeepMind
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolNeel Kovelamudi is a research engineer at DeepMind focused on foundational work toward AGI, drawing on six years of ML and software engineering experience across Google Brain, Keras, and Gemini production infrastructure. He helped implement Keras 3.0 and led critical saving/export and serialization work that enables millions of users to deploy models, and played a key role in launching Gemma—the first open-source multi-backend LLM—alongside DeepMind, NVIDIA, and HuggingFace. At Google he built production systems for Gemini model training, checkpointing, and multi-modal evaluation, earning a T-rating in recognition of top-tier engineering impact. An active open-source contributor, his code and documentation improvements in keras and TensorFlow focus on practical training patterns, serialization robustness, and TFLite export reliability. Trained at Johns Hopkins and UC Berkeley, he combines research rigor with production-grade engineering and a knack for turning complex serialization and export edge cases into reliable APIs.
6 years of coding experience
5 years of employment as a software developer
Master's degree Electrical Engineering and Computer Science, Master's degree Electrical Engineering and Computer Science at University of California, BerkeleyJohns Hopkins University
English, Chinese
Github Skills (13)
serialization10
keras10
tflite10
machine-learning10
data-serialization10
deep-learning10
tensorflow10
python10
back-end-development10
image-classification9
unit-testing9
deeplearning-ai8
neural-network7
Programming languages (4)
Github contributions (5)
keras-team/keras
Sep 2022 - Jan 2023
Deep Learning for humans
Role in this project:
Back-end Developer
Back-end DeveloperContributions:29 reviews, 21 commits, 20 PRs in 4 months
Contributions summary:Neel primarily contributed to the Keras library, focusing on refactoring and updating the serialization process. Their work involved moving code, adding support for object serialization, and enhancing existing serialization functionality for various components. This involved significant code changes related to class configurations, object tracking, and backwards compatibility for different saving formats. They also added tests to cover the changes.
pythondata-sciencedeep-learningneural-networksmachine-learning
keras-team/keras-io
Sep 2022 - Sep 2022
Keras documentation, hosted live at keras.io
Role in this project:ML Engineer
ML EngineerContributions:36 reviews, 1 commit, 19 PRs in 1 day
Contributions summary:Neel's commits primarily involve modifying and adding examples related to the Keras framework, specifically demonstrating the "Trainer pattern" for custom training steps and creating a guide for saving and serialization. They also updated the example for the EANet, MLP image classification, and MixUp examples to be compatible with Keras 3. The user's contributions are focused on enhancing the documentation and examples to provide insights on training techniques and model saving.
deep-learninghostedmachine-learningtensorflowkeras
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.