Preintegrated wheel odometry with online slip estimation for offroad autonomous ground vehicles.

Structure from Motion implemented in the new programming language Mojo.

A neural radiance field using fisheye cameras and online pose estimation.

Structure from Motion implemented in python from scratch.

Full robotics stack include estimation, planning, & control for an AUV in HoloOcean.

Inverse Kinematics and Trajectory Generation for the EezyBotArm Mk2 built on ROS and a raspberry pi.

Improving the extended Kalman filter using deep learning.

HoloOcean is a high-fidelity underwater simulator forked from Holodeck and built on Unreal Engine.

I co-founded SiPANN (Silicon Photonic Artificial Neural Networks). It's a python package used to accelerate the design of silicon photonics.

For a data analysis course, I cleaned and analyzed NICU EEG (brain scan) data for trends about how and when premature babies have seizures, and drew various conclusions.

As an application of optimal control theory, I optimized a weighted combination of fuel and time to get a car from point A to B over an arbitrary surface.

I run Printed Practicals, a 3D printing business focusing on selling mounts for Google Home products. We have sold over a 1000 orders to all around the world.

Using a Convolutional Neural Network, sort oreos as good, bad or ugly

Using various machine learning techniques and a public recipe reviews dataset, recommend similar recipes to reviewers.

The InEKF is extension of the Kalman Filter built upon Lie Groups with guaranteed asymptotic stability. We implement a few examples.

To practice C++ coding paradigms, I implemented various portions of the popular python Sklearn package in C++.

Using NFC tags & a Raspberry Pi, I implemented a diy Vinyl that plays spotify albums on google homes.

Using Aruco tags & OpenCV, implemented a simple Pokemon Augmented reality program.

Derivation & Simulation Results of the InEKF for use with underwater sensors such as an IMU, DVL and pressure.

Simple implementation of iSAM using Aruco tags as markers and gtsam as the backend.

We use a Kalman Filter and calibrated stereo cameras to track a ball in 3D.

Use PDF writeups (often written in LaTeX) as Github READMEs using Github Actions.