Long informal bio

Often times when I introduce myself I forget a few things I worked on, so this is in part an exercise for myself. I also think this is cooler than a resume if you want to know about me.

Before undergrad (-2020)

I was always interested in science + computers as a kid. I started thinking about AI because I wanted to send computers into deep space for space exploration instead of humans.

I also wanted to hack into video games. Since I grew up kind of poor, I started buying j-tagged xbox 360s and selling mod menu access. I also modded 3DSs for fun. This was my first intro to CS after "hello world".

My interests became a little more practical after seeing some of the first Tesla autopilot demos. It amazed me (and still does) that you could learn how to control a real car by training a computer on collected data.

Following my newfound interest I joined the VEX Robotics team at my high school in my junior year where I learned the basics of robotic mechanics, control, and electronics. Some fun things I did during VEX:

• Went to worlds
• Won a few awards from local competitions
• Built a teleoperation-replay program to quickly create new autonomous routines
• Built an automatic target shooter using classical computer vision techniques
• Went to a Google Signature Event competition and visited Silicon Valley for the first time
• Led a team and designed+built my first robot end to end

• Progressed through all the virtual rounds, went to the final competition at MIT
• Met some astronauts at MIT
• Spoke to astronauts aboard the international space station live
• Ran my autonomous code on the SPHERES sattelites on the ISS
• Said "yeet" to an astronaut on the ISS for a dare

Undergrad (2020-2023)

I joined Purdue a bit jaded at robot learning. Much of the classical autonomy work for VEX Robotics did not interest me. I wasn't exposed to any ideas in robot learning outside of reinforcement learning, which I was initially dismissive toward since it didn't solve any real-world robotics problems that I knew of. Purdue CS also didn't have any Robotics faculty at this time. A lot of the robotics work I found interesting at the time did some classical planning on top of a state representation from computer vision models.

So, my interests turned toward Computer Vision for a bit. I joined the TensorFlow Model Garden lab at Purdue under Dr. James Davis. The lab worked on reproducing various YOLO models into TensorFlow's open source model garden. At the lab I :

• Increased the average precision of our YOLOv4-tiny model from 16% to 21.2%
• Created a TF-Record generator from the Pix3D dataset for the development of a 3D mesh predictor model
• Found a bug within TensorFlow where evaluating popular models on TPUs causes out-of-memory crashes

I also did a lot of hackathons.

During my Freshman year summer I interned at Pacific Northwest National Laboratory, where I helped benchmark frameworks that did sparse matrix operations to guide design decisions for a Chemistry DSL.

Sophomore Year

Purdue has one of the best VEX U Robotics teams, and VEX Robotics reached out to us to help develop some things for their AI competition. I led a team of 6 and we:

• Annotated a bunch of game object data
• Created a game object detector + localizer using RGBD cameras
• Made a sim and RL based controller to play the game (it didn't work well lol)
• Developed a protocol to interface between the VEX microcontroller and a Jetson
• Open sourced everything and wrote a report

Junior Year

Before going back to school, I took a fall SDE internship at Amazon. Because of some funny quirks in recruiting, I was one of about 3 dozen interns directly reporting to a Senior Principal Engineer. Due to my interest in deep learning, I had the fortune of being placed on an applied science team, doing applied science work, as a SDE intern. I worked with Dr. Jangwon Kim and Dr. Pragyan Mishra on Healthcare NLP applications and gave a presentation to the senior leadership at Amazon's healthcare org. From what I know, there were plans to turn my intern project into a product, but I can't find it anywhere :)

During my internship at Amazon I visited a friend at Stanford who started working on visuomotor policy research with Professor Chelsea Finn and was subsequently introduced to a lot of great robot learning research coming out of Everyday Robots (rip) and collaborators. This was the largest inspiration to pursue my current work in robot learning.

I also stumbled upon the Amazon Alexa Prize challenges while I was an intern and thought it was interesting. I wrote a proposal with my good friend Jinen Setpal, submitted it under a grad student's name (we were both undergrads at the time and were uneligible), and miraculously won $250k to work on multimodal ai for Amazon Alexa.

Returning to Purdue, I joined the CoRAL Lab under Professor Ahmed Qureshi. I started a research project adding gaze control to visuolinguomotor policies so they could disambiguate language prompts. The control is conditioned on eyesight gaze and language, so the model performs well even when the control prompt is ambiguous, such as "move that block to the top right". This project was essentially an extension of Interactive Language. I had decent results in sim, but unfortunately was never able to close the sim2real gap!

During my junior year I started ML@Purdue, and I regard this as my greatest non-technical achievement so far. I started it to unite ML students from Purdue's three schools of computing, Purdue CS, Purdue ECE, and Purdue Polytechnic into one club. I'm very proud of the community and happy to be a part of so many student's first intro to machine learning :)

For my last internship I joined my Amazon manager's new startup, Armada AI. I was one of the first members of the AI team and did some of their first demos for computer vision tasks on their edge ai platform.

Senior Year

I mostly continued my work at CoRAL, Armada, and ML@Purdue for my last semester.

After undergrad (2024-)