Naren Dasan

About me

Hi! My name is Naren. I am from Colorado, did my undergrad in the cornfields of central Illinois and after a stint in San Francisco working on the development of self driving cars, I am now a graduate student at USC.

I've worked on topics ranging from climate modeling, high performance computing, user interface design, computer vision and embedded and operating systems but since I was in high school I have most passionate about thinking about and developing the ways one or many people and one or many robots interact with each other (i.e. human robot interaction, swarm robotics, multi-agent systems, and artificial intelligence). Hence a good amount of my academic research is centered around these ideas.

I am also an advocate for Free and Open Source Software and Hardware. During my undergrad, I, along with a couple close collaborators worked on extending the ideas of the hackathon craze of the 2010s to address getting more students involved in the open source community. This resulted in HackIllinois becoming (to my knowledge) the first open source collegiate hackathon. Growing from a small pilot program to 1000+ attendees a year, over the last 4 years 100s of Pull Requests, a few releases, and a whole lot of cool new open source projects and forks were created at Hackillinois, a lot done by first time open source contributors being mentored by core maintainers of projects like Debian, Rust, WebAssembly, Python, Julia and more.

In my free time I'm an avid mountain biker (and road biker too I guess) as well as like to play tennis, soccer and rugby (have worn the jersey for both Illinois and USC). I was chair of the UIUC Chapter of ACM from 2016-2017, ACM@UIUC being the nations largest ACM student chapter and during my time there I created quite a few projects that ACM now maintain.

I work at NVIDIA developing perception and other deep learning components for autonomous vehicles within the constraints of the resource limited environment of a car. I therefore use TensorRT a lot and ended up creating the TensorRT Python API and worked on the open sourcing effort for TensorRT. I also created and am the lead maintainer of TRTorch now Torch-Tensorrt, a compiler for PyTorch that targets NVIDIA GPUs via TensorRT. You can see a hint of the perception work we do by looking at NVIDIA's DevBlog.

At USC, I work at the ICAROS Lab lead by Prof. Stefanos Nikolaidis. My research centers focuses on a couple aspects of developing human compatible robots. One of my main thrusts is in enabling robots to quickly learn the workflow preferences of their human counterparts in collaborative tasks. Another of my thrusts is in leveraging quality and diversity algorithms to train more robust agents.