Naren Dasan
Sr. Automotive Deep Learning Solution Architect at NVIDIA
Boulder, Colorado, United States
Summary
🤩
Rockstar🎓
Top SchoolNaren Dasan is a Sr. Automotive Deep Learning Solution Architect with over a decade of experience applying embedded systems, robotics, and computer vision to real-world products, currently leading Torch-TensorRT engineering at NVIDIA to accelerate PyTorch inference on NVIDIA GPUs. He blends deep research experience from UIUC in surface reconstruction and robotic grasping with hands-on systems work in C++, Python, and Go/Ruby to ship production-ready, low-power perception stacks for autonomous vehicles. Naren is an active open-source contributor and MLOps engineer, improving PyTorch→TensorRT tooling and build flexibility—work that increases the set of ops convertible to high-performance runtime. He has a track record of bridging research and deployment: from creating TensorRT Python APIs as an intern to driving compiler-level inference optimizations today. Based in Boulder, he pairs academic rigor with pragmatic engineering, often tackling the gritty build and dependency issues that make ML systems reliably deployable.
12 years of coding experience
8 years of employment as a software developer
Bachelor’s Degree Computer Engineering, Bachelor’s Degree Computer Engineering at University of Illinois Urbana-Champaign
Stackoverflow
23reputation
1kreached
0answers
1question
Github Skills (15)
tensorrt10
pytorch10
c-language10
cprogramming-language10
python10
build-automation10
machine-learning9
deep-learning9
cuda7
gdb6
linux6
binarydiff6
dwarf6
dockers5
docker5
Programming languages (20)
Github contributions (5)
pytorch/TensorRT
Mar 2020 - Jan 2023
PyTorch/TorchScript/FX compiler for NVIDIA GPUs using TensorRT
Role in this project:
ML Engineer
ML EngineerContributions:17 releases, 1527 reviews, 1343 commits in 2 years 10 months
Contributions summary:Naren's commits primarily focused on modifications to the documentation and the execution phase of the PyTorch/TorchScript/FX compiler for NVIDIA GPUs using TensorRT. They updated execution phase documentation and added new evaluators. The contributions show that the user worked on making a larger set of operations convertable by creating new evaluators for basic math operations such as aten::add and aten::mul.
cudapytorchjetsonnvidiadeep-learning
NVIDIA-AI-IOT/torch2trt
Jul 2019 - Jul 2019
An easy to use PyTorch to TensorRT converter
Role in this project:MLOps Engineer
MLOps EngineerContributions:10 commits, 3 PRs in 2 days
Contributions summary:Naren focused on enhancing the build process and dependency management for the PyTorch to TensorRT converter. They implemented the ability to specify custom library locations for CUDA, PyTorch, and TensorRT, directly improving the build flexibility. The user also added a `clean` command for removing build artifacts and integrated the build process with `setup.py develop` to streamline the development workflow. Moreover, they addressed compilation issues by adding the `-std=c++11` flag.
pytorchconverterjetson-nanojetson-xavierinference
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.