Layali R
Deep Learning Architect at NVIDIA
Issaquah, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolLayali R is a Deep Learning Architect with eight years of experience scaling large language models to run efficiently across thousands of GPUs, currently driving GPT-3 performance work on NVIDIA’s MLPerf-training team. She combines deep hardware and systems expertise—from CPU microarchitectures at Qualcomm and IPU enablement at Microsoft—to optimize model and data parallelism, network topologies, and memory placements for production-scale training. Her background in cycle-accurate simulation and fault-tolerant research (PhD work) gives her a rare ability to profile, model, and pinpoint low-level bottlenecks that impact end-to-end ML throughput. Known for proposing hardware-aware sparsity and system-level enhancements (including patent submissions), she bridges research, performance engineering, and practical deployment to squeeze out real-world gains. Based in Issaquah, WA, she thrives at the intersection of architecture and ML systems engineering, turning profiling insight into scalable training solutions.
8 years of coding experience
10 years of employment as a software developer
Ph.D, Computer Engineering, Ph.D, Computer Engineering at The University of British ColumbiaM.Sc, Computer Engineering, M.Sc, Computer Engineering at The University of Calgary
Github Skills (29)
transformers10
speaker-diarization10
speech-to-text10
machine-translation10
speech-synthesis10
multimodal10
large-language-models10
nmt10
text-normalization10
transformer-models10
deep-learning10
generative-ai10
speaker-recognition10
natural-language-processing10
asr10
Programming languages (1)
Github contributions (5)
layalir/NeMo
Mar 2023 - Apr 2024
NeMo: a toolkit for conversational AI
Contributions:1 PR, 21 pushes, 18 branches in 1 year
layalir/TransformerEngine
Jun 2023 - Feb 2024
A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit floating point (FP8) precision on Hopper GPUs, to provide better performance with lower memory utilization in both training and inference.
Contributions:6 pushes, 3 branches in 7 months
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.