Pavani Majety
Deep Learning Engineer, DL Compilers at NVIDIA
California, United States
Summary
🤩
Rockstar🎓
Top SchoolPavani Majety is a Deep Learning Engineer at NVIDIA with eight years of experience building high-performance inference solutions and compiler-level optimizations for DL workloads. Based in California, she focuses on DL compilers and code generation, translating research-grade models into production-ready, memory- and compute-efficient deployments. Her open-source contributions include performance work on vLLM’s FlashInfer backend—adding FP8 KV cache support, sliding-window inference, and targeted bug fixes—demonstrating a practical emphasis on throughput and numeric-efficiency for large language models. Trained at VIT and the University of Michigan, she blends strong academic foundations with hands-on systems engineering to squeeze latency and memory out of real-world inference stacks.
8 years of coding experience
Master of Science - MS, Master of Science - MS at University of MichiganBachelor of Technology (B.Tech.), Bachelor of Technology (B.Tech.) at Vellore Institute of Technology
Tamil, French, Hindi, English, Telugu
Stackoverflow
11reputation
236reached
0answers
2questions
Github Skills (14)
llm10
cuda10
pytorch10
deep-learning10
inference10
pytest9
testing9
transformer8
file-handling6
eclipse6
socket6
command-prompt6
testng6
java6
Programming languages (6)
Github contributions (5)
vllm-project/vllm
Jun 2024 - Apr 2025
A high-throughput and memory-efficient inference and serving engine for LLMs
Role in this project:
ML Engineer
ML EngineerContributions:27 reviews, 14 PRs, 72 comments in 9 months
Contributions summary:Pavani's commits primarily involve enhancing and testing the FlashInfer backend within the vLLM project. They focused on integrating and enabling FP8 (float8) KV Cache with the FlashInfer backend, including both prefill and decode operations. The user contributed to adding tests for the FP8 KV Cache and addressing related bug fixes, showcasing their work in performance optimization and efficient inference techniques for LLMs. They also added sliding window support with Flashinfer, and addressed several bugfixes related to K scale and V scale.
amdcudadeepseekgpthpu
pavanimajety/tensorflow
Jul 2022 - Mar 2023
An Open Source Machine Learning Framework for Everyone
Contributions:124 pushes, 14 branches in 7 months
pythondata-sciencedeep-learningmachine-learningframework-learning
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.