Find more about my experience and work related projects on LinkedIn

Project Portfolio

Intel® Edge AI for IoT Developers Nanodegree

Scholarship Winner

Udacity Robotics Software Engineer Nanodegree

Term 1

Project 1- Pick and Place (Kinematics)
Derive and implement inverse kinematics to guide 6-DOF robot arm to target pick and place locations
Report/Writeup —Repository

Project 2- Pick and Place (Perception-Traditional)
Develop a 3D perception pipeline using RGB-D camera to generate grasp targets for a pick and place operation
Report/Writeup —Repository

Project 3- Follow Me (Perception-Deep Learning)
The central problem this project addresses is semantic segmentation of an image. In our scenario, we have a quadcopter with a camera that we would like to follow a specific person. A dataset of example images has been provided as well as ground truth segmentation of these images. The proposed solution implements supervised learning to train a fully convolutional neural network to perform semantic segmentation on images to classify our target person, other people, and the background. The effectiveness of the solution is measured using Intersection over Union of the segmentation compared to the ground truth.
Report/Writeup —Repository

Term 2

Robotic Systems Deployment (Neural Network Training and Inference)
two deep neural networks are trained to perform object classification utilizing the Nvidia Digits software framework. These models are trained on a dataset provided as part of the Udacity Robotics Software Nanodegree and a custom dataset. A discussion of each implementation covers problem background information, data acquisition, model selection, training and inference, model results, and possibilities for future work.
Report/Writeup —Repository

Where Am I? (Localization)
This project focuses on the implementation of localization for a mobile robot. Adaptive Monte Carlo Localization is implemented using 2D LIDAR scan data. This algorithm is also compared with Kalman Particle Filters. Two robot models are compared regarding physical configuration and optimization of the AMCL parameters.
Report/Writeup —Repository

Map My World (Simultaneous Localization and Mapping)
Mobile robot navigation is often dependent on an accurate map of the environment the robot is operating in. While in some cases, a static map may be sufficient if the environment will not change. However, most environments change periodically and so to accurately and efficiently navigate through a changing environment the robot must have the ability to create a map. There are several different methods to implement mapping, this project focuses on implementing The goal of this project is to create a ROS package that successfully implements environment mapping on a mobile robot in simulation. The robot is tested in two simulated environments
Report/Writeup —Repository

Pick and Place with Deep Learning (Deep Reinforcement Learning)
The purpose of this project is to implement an agent, develop rewards functions, and tune hyperparameters to train an agent using deep reinforcement learning to control a robot arm in simulation. The goal is to achieve 90 percent accuracy with any part of the robot arm hitting the target object and 80 percent accuracy with the gripper base hitting the target object.
Report/Writeup —Repository

Home Service Robot (Path Planning and Navigation)
This repository contains code for creating a robot that can perform tasks in a home service setting. The example implementation shown here uses the example of item pickup and delivery. The end result is a robot that can navigate through a previously mapped environment to a pick up location to retrieve an item and then navigate to a delivery location to deliver that item in a simulated environment. The Home Service robot uses markers inside RVIZ to represent the real world item.
Report/Writeup —Repository

Udacity Machine Learning Engineer Nanodegree

Completed December 2016 Project code and reports from Udacity’s Machine Learning Nanodegree

Project 1-Predicting Housing Prices in Boston (Modeling and Validation)
Report/Writeup —Repository

Project 2-Determining Student Intervention (Supervised Learning)
Ipython Notebook —Repository

Project 3-Determining Customer Segments (Unsupervised Learning)
Ipython Notebook —Repository

Project 4-Training a smartcab to drive (Reinforcement Learning)
Report/Writeup —Repository

Capstone- Navigating a Maze (Robot Motion Planning)
Report/Writeup —Repository