Trevor Morris
Senior Deep Learning Software Engineer at NVIDIA
Irvine, California, United States
Summary
🤩
Rockstar🎓
Top SchoolTrevor Morris is a Senior Deep Learning Software Engineer with a decade of experience optimizing ML systems for GPUs and compilers, currently working on TensorFlow frameworks at NVIDIA. He combines deep learning research roots from UCSB (BS/MS) with hands-on production work at NVIDIA and AWS, where he integrated TVM and TensorRT to deliver 2x–10x inference speedups across cloud and edge GPUs. Trevor is a prolific open-source contributor—impacting TensorFlow, TVM, IREE, and NVIDIA’s OpenSeq2Seq—and has added BF16/GPU kernels, TensorRT integration, and collective ops support that enabled real-world deployment of quantized and high-performance models. Comfortable across the stack from ML research (GAN-based super-resolution) to compiler backends and CUDA runtime tweaks, he repeatedly bridges algorithmic insight with low-level systems engineering. Based in Irvine, CA, he brings a rare mix of academic rigor and production-scale compiler/GPU optimization experience that accelerates both research prototypes and mission-critical inference.
10 years of coding experience
5 years of employment as a software developer
Bachelor’s Degree, Computer Engineering, 3.77 GPA, Bachelor’s Degree, Computer Engineering, 3.77 GPA at University of California, Santa Barbara
Github Skills (34)
tvm10
python10
image-classification10
tensorrt10
gpu-programming10
relay-modern10
machine-learning10
inference10
mlr10
compiler-compiler10
deep-learning10
tensorflow10
optimisation10
ccl10
compiler10
Programming languages (6)
Github contributions (5)
apache/tvm
Nov 2019 - Jul 2021
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
ML Engineer
ML EngineerContributions:98 reviews, 52 commits, 59 PRs in 1 year 8 months
Contributions summary:Trevor contributed to the TensorFlow frontend within the TVM project, focusing on implementing support for new TensorFlow operations. Their work included fixing issues with existing operations like `GatherV2` and implementing converters for new operations, such as `max_pool2d_with_indices` and `CombinedNonMaxSuppression`, and addressing issues related to padding. The user's contributions also involved improvements to existing converters to handle dynamic batch sizes and avoid unnecessary type casts, showcasing expertise in TensorFlow and relay.
metalvulkancompilertensoropencl
tensorflow/tensorrt
Nov 2018 - Jul 2019
TensorFlow/TensorRT integration
Role in this project:ML Engineer
ML EngineerContributions:7 commits, 9 PRs, 7 comments in 7 months
Contributions summary:Trevor primarily focused on enhancing the image classification example within the TensorFlow/TensorRT integration. Their contributions include adding an inference script and documentation, updating links, and improving the object detection example by utilizing the new TensorRT API for conversion and calibration. They also simplified model download procedures and optimized the code by using tf.variable instead of tf.constant for synthetic input.
nvidia-dockerdeep-learningtensorflowtensorrt
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.