About

For the past ~2.5 months I have been independently working on projects to build ML engineering skills in performance optimization and robotics data processing. The projects below are from this time-period as well as a few that I completed during my free-time at Tesla.

‍Diffusion Policy Inference Optimization
‍I used a combination of custom GPU kernels for 1D Convolution, diffusion-based denoising, Group Normaliation + Mish activation, and CUDA graphs to reduce inference time for the convolutional U-Net by 3.4x. I documented in work in detail here.

Code and more detailed description for the projects below can be found here.
‍
‍Object Grasping with 6-DoF Robot Arm
‍I used Soft-Actor Critic (RL algorithm) to train a policy in simulation (Pybullet) to learn how to pick up blocks. The goal with this project was to transfer the learnings in simulation to a calibrated test-environment with a real Dofbot but unfortunately servo controllers on my Dofbot died before I got around to trying to make sim2real work. I started out with RL because I thought the connections between TD-error and dopamine spikes in monkeys (and presumably humans) were fascinating! After spending weeks getting a policy working with RL, I trained a policy in a weekend using imitation learning and an oracle which outperformed the one trained with SAC. Given how incredibly dextrous humans are, I think focusing on scaling imitation learning is probably the right idea (for personal projects and for the object manipulation problem in general).

‍Neural Network Visualizer
‍I implemented an MLP + backprop from scratch using NumPy, trained it to classify cat images, and used Matplotlib to visualize the training process.

‍Real-Time Robot Arm Control Using Pose Estimation
‍This project involved using a pre-trained pose estimation model (implemented via MediaPipe) to control a Kuka robot arm in a virtual PyBullet environment. While not particularly technically challenging beyond utilizng some simple 2D kinematics, this project was a fun way to implement a useful AI model and learn the PyBullet API in the process.

Tesla was a dream company of mine to work at in undergrad and I was incredibly excited when I got the email to interview with the HV Devices team my Junior year at UF. I ended up spending most of my professional career at Tesla, on HV Battery and adjacent teams, until I left in June 2023. I was lucky to work with lots of very smart and kind people there who taught me valuable lessons about engineering for high-volume and dealing with real-world complexity. Below are some highlights from the different positions I held during my time there.

Sept 2019 to December 2019 - Mechanical Design Engineering Intern on High Voltage Devices‍

Developed a custom Hall-effect based current sensor optimized for detecting fault currents up to 25kA. Leveraged Ansys Maxwell to enhance busbar geometry and shape the resulting magnetic field from fault current, achieving an optimal signal-to-noise ratio while minimizing material removal from the busbar cross-section. Validated simulation results with a prototype, demonstrating accuracy within 5%.
Simultanouesly developed a complete injection molded plastic assembly for Semi HV Battery within a tight 12-week timeline, successfully kicking off tooling before end-of-internship without prior experience in designing plastic components.

Aug 2020 to Dec 2020, Mechanical Design Engineering Intern on Battery Structures

Designed a stamped structural member of the Model Y Structural Battery. Worked with partner team (CAE) to simulate the impact this member had on vehicle crash safety and iterated to reduce likelihood of passenger injury during front crash cases
Provided factory support during low-volume prototype builds of Model S Plaid. Collaborated with NPI (New Product Introduction), Production Associates, and Process teams to debug design issues with new battery architecture

May 2021 - July 2023, Mechanical Design Engineer on Battery Design

One of two design engineers responsible for a ground-up redesign of a critical subsystem in the Model Y Structural Battery (8 week effort, start-to-finish)
Sole design owner for HV Sense & Control sub-assembly in the Model Y Structural Battery. Collaborated cross-functionally with partner teams internally and external vendors to carry design from scratch to high-volume production
Developed proof-of-concept for innovative lithium-ion battery safety mechanism. As part of this project, I built out prototyping infrastructure to circumvent time-consuming live thermal runaway test pipelines. This involved a representative and portable battery module, in-house steel heat-treating, 3D printed metal forming dies, and more (under NDA).

2016 to 2017 - Brakes & Solar Array Lead, Sponsorship Lead
I got involved with Solar Gators in my freshman year and was technical lead for Brakes and Solar Array, while also leading our sponsorships strategy. Solar Gators pulled off an extremely unlikely feat by qualifying for the FSGP 2017 on our first try, with a budget smaller than the price of most teams motors! Seeing our incredibly janky car, full of blood/sweat/tears (actually), make it up Turn 1 at Circuit of the Americas is one of the hypest moments of my life! I went into this team knowing nothing about hydraulic brakes or monocrystalline silicon solar cells and since the team was brand-new, so did few others. This experience taught me how much you can learn in a short period of time if you’re just willing to keep at it!

Our first car, Torch

2017 to 2018 - Treasurer, Aero-Body Lead
In my Sophomore year I was on the Exec board for Solar Gators as Treasuer and also sub-system lead for our first ever carbon fiber aero-shell. As Treasurer I worked to expand our total budget for Cielo (our second car) to $125k ($50k cash + $75k in-kind donations). I also secured sponsorships to enable manufacturing of the aero-shell that I designed using Solidworks (CAD) + Star-CCM (CFD). It was really cool to see something I made in CAD turn into a very expensive outer-shell for a solar car in ~6 months!

CAD render of aero-body for Cielo

Carbon fiber parts fresh out of the mold!

Cielo + Team + me in the car!

‍

2018 to 2019 - President, Chief-Engineer
My Junior year I was President of Solar Gators and chief-engineer for all design & manufacturing activities. During this year we made many improvements to Cielo I and also began design for our most ambitious car yet, Sunrider. Technical management of a team of ~25 interdisciplinary engineers while keeping up with school was a great forcing function for growth. I was also very proud that Cielo II performed many times better in FSGP 2019 than the original version in 2018.

Cielo II raycing at Circuit of the Americas in Austin

‍

2020 to 2021 - Advisor, Last-minute Rolling Chassis Lead
COVID hit Solar Gators hard - its pretty hard to build solar cars remotely! On-boarding new members and making progress on Sunrider (our third car) was very tough for the team. I mostly remained in an advisory role until 2 months before graduation, when I became very worried that very little of Sunrider physically existed and that Solar Gators would fall apart after the current seniors left. I, along with a handful of others, led a blitz to get Sunrider to rolling-chassis status and we worked non-stop to get to a fully functional powered chassis (technically a week after graduating). I am super glad that we did this as the new crew of Solar Gators kept the momentum going after we left. I was obviously incredibly thrilled that Sunrider got 1st place in its second race (FSGP 2023)!

Sunrider - definitely our most aesthetic car

As I write this up, I can’t help but reflect on how awesome its been to be a part of something much bigger than me and hope Solar Gators continues to be a proving ground for new generations of young engineers!

About Me

Experience