Sravya Popuri
Research Manager, Llama Pretraining at Meta
Menlo Park, California, United States
Summary
🤩
Rockstar🎓
Top SchoolSravya Popuri is a research-focused engineering leader with five years of experience building foundational ML systems for speech and multimodal models, currently managing Llama pretraining at Meta. She previously led research engineering for FAIR’s Seamless suite for speech translation and has hands-on experience integrating advanced architectures (e.g., conformers, Simul ST) into the popular fairseq toolkit. Her background spans applied machine learning roles at Microsoft and Meta and a master’s in Intelligent Information Systems from Carnegie Mellon. Based in Menlo Park, she combines production-savvy software engineering with research-driven model design, bridging pretraining architecture exploration and deployment considerations. An interesting through-line in her work is elevating speech-to-speech translation latency and robustness while shipping at scale for global communication products.
5 years of coding experience
8 years of employment as a software developer
Bachelor’s Degree Computer Science and Engineering, Bachelor’s Degree Computer Science and Engineering at International Institute of Information Technology Hyderabad (IIITH)Master’s Degree Masters in Intelligent Information systems, Master’s Degree Masters in Intelligent Information systems at Carnegie Mellon University
English, Hindi, Telugu
Github Skills (11)
pytorch10
machine-learning10
artificial-intelligence10
integrate10
conformer10
python10
integrations10
data-integration10
fairseq10
model-optimization10
nlp9
Programming languages (3)
Github contributions (5)
facebookresearch/fairseq
Feb 2021 - Oct 2022
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Role in this project:
ML Engineer
ML EngineerContributions:30 reviews, 33 commits, 19 PRs in 1 year 7 months
Contributions summary:Sravya's commits primarily involve integrating and modifying existing machine learning models within the `fairseq` framework, a sequence-to-sequence toolkit. Their work includes integrating a Simul ST model, updating models for augmented memory, and integrating conformer models, suggesting a focus on advancing speech-to-speech translation capabilities. The changes range from minor code adjustments to incorporating new architectures, showcasing their involvement in model development and adaptation within the project's core functionalities. They also add latency metrics for the inference script.
pytorchnlpsequencepythontransformer-architecture
sravyapopuri388/fairseq
May 2022 - Sep 2022
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Contributions:7 PRs, 51 pushes, 1 branch in 3 months
nlpsequencepythonmachine-learningfacebook
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.