Ekagra Ranjan
Engineer at Cohere
Canada
Summary
👤
Senior🎓
Top SchoolEkagra Ranjan is an engineer with nine years of experience building and deploying deep learning systems across graphs, vision, and language, currently focusing on LLM inference and efficiency at Cohere. He combines a research mindset—publishing at AAAI with work cited 200+ times—with hands-on engineering at Microsoft where he doubled relevance in knowledge-graph recommendations and shipped low-latency custom message-passing algorithms. An active open-source contributor, he has improved major libraries including PyTorch Geometric and TorchVision and contributed to Hugging Face and vLLM, often enhancing model flexibility and usability (e.g., multi-size inputs, pooling layers, and affine transforms). Comfortable across CUDA, Triton, Python and C++, he also bridges production and research by building CI/CD, migrating Spark pipelines to Azure ML, and prototyping distributed graph learning. Outside work he pursues open-source projects and a surprising taste for stand-up comedy, reflecting a practical but curious approach to problem solving.
9 years of coding experience
3 years of employment as a software developer
Bachelor of Technology - BTech, Electrical, Electronics and Communications Engineering, Bachelor of Technology - BTech, Electrical, Electronics and Communications Engineering at Indian Institute of Technology, Guwahati
Github Skills (43)
text-generation10
data-generation10
music-generation10
pytorch10
language-model10
python10
pre-trained-model10
model-driven10
avatar-generation10
machine-learning10
document-generation10
nlpjs10
mask-rcnn10
model-building10
deeplearning-ai10
Programming languages (8)
Github contributions (5)
pyg-team/pytorch_geometric
May 2019 - Jun 2020
Graph Neural Network Library for PyTorch
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:54 commits, 28 PRs, 29 comments in 1 year 1 month
Contributions summary:Ekagra's contributions primarily focused on improving the PyTorch Geometric library. They addressed code quality issues by fixing typos and adding missing variables within the codebase. Furthermore, they enhanced the documentation and example code, and added new features like a pooling layer, and added support for the GCN and GraphSAGE models with Jumping Knowledge.
pytorchgraph-convolutional-networksgeometric-deep-learningdeep-learningneural-graph
dotnet/TorchSharp
Oct 2021 - Oct 2021
A .NET library that provides access to the library that powers PyTorch.
Role in this project:ML Engineer
ML EngineerContributions:2 reviews, 14 commits, 6 PRs in 5 days
Contributions summary:Ekagra's commits primarily focused on improving documentation related to image transformations within the TorchSharp library, including Gaussian blur, random rotation, random resized crop, perspective transforms, padding, and resizing. They also made changes to the Functional interface of TorchSharp, refactoring and optimizing code. These changes suggest a focus on refining the usability and documentation of core image processing functions within the library.
pytorchnet-librarydotnetpythonpowers
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.