Currently working on an advanced AI tool designed to extract and analyze relevant news data from various websites, with a specific focus on sentiment analysis related to companies and stocks. This innovative solution aims to provide traders and investors with valuable insights into market sentiment, enabling more informed decision-making in the dynamic world of financial markets.
Utilized transfer learning to retrain a convolutional neural network which is adept at object detection. Transfer learning is a technique to improve the learning process of a new task through the transfer of knowledge from a related task that has already been learned. This will be applied to a Convolutional Neural Network, a class of AI network that is commonly used to analyze visual imagery. My dataset consists of traffic images where I identified and classified different types of cars. The first task is selecting an optimal model for this task. I ended up selecting the YOLOv5 model, a well known and widely used model The model was trained on the dataset of 15000 images. Once a retrained model was achieved, I was able to run it on the test dataset for roughly 50 epochs to produce labels of its own, labels which define bounding boxes for objects, their classification and confidence rating of that prediction from the model. I ended up being rank 1 at the time of writing this report on CLP with a MaP of .6697
This application is a web E-Commerce platform where users can create an account, browse & search through the inventory using a catalog with infinite scroll, make purchases, & leave reviews on products. The seller can also access an admin panel where they can track inventory stock, adjust prices and add new products. This app was created using Ruby on Rails for the backend along with the Devise & Doorkeeper gem for Oauth2 authentication, the front end was designed using bootstrap with responsiveness in mind.
This project was something I worked on during my spare time for a client of mine. I theorized and developed it from the ground up.This program monitored the network status of user inputted IP addresses WAN connection, routers, and VPN tunnel. It logged latency and generated analysis and statistics for each location. I have plans to display a real-time graph of the latency within the GUI in a drop down menu that will also display additional relevant information such as average, highest, and lowest latency. Once deployed this proved to be an invaluable networking tool that my client greatly benefited from. This project was completed using Python and the TKinter library.
This application was made to familiarize myself with Oauth2, seeing as how it is an industry standard protocol for authorization, I wanted to have a deep understanding of the technology. I made an app that authenticated the user with Oauth 2.0 using the websites Facebook, Google, Github, and my own E-Commerce app as providers.
This project was something I worked on for a client of mine. I theorized and developed it from the ground up. this program Automates the process of making MAC address based IP reservations in a Fortinet 60E Router and Fortimanager using Python, Paramiko (a python module used for ssh connections), and the Fortinet CLI. Making, modifying, and removing DHCP reservations was a very comman task performed at my clients business. This required loading a clunky web application supplied by Fortinet and performing various boilerplate tasks through the Fortinet GUI. This cmd based approach eliminated the sluggish behaviour due to the bloated Fortinet GUI and eliminated all of the boilerplate tasks, just requiring the bare essential information from the user. Once developed this vastly improved work efficiency at my clients site.
I used a AT&T IoT Development Sensor Board (AT&T NXT FRDM-K64F Development Board with an Arduino WNC Shield) equipped with 2 LTE antennas using wireless cellular conntection powered by the AT&T network as well as temperature and humidity sensors and a accelerometer, to build a streaming application. The app was programmed in AT&T Flow and the data from the board was pushed to AT&T M2X. From there I used PubNub to publish the data and then Microsoft Bi to subscribe to the data channel and live stream the data from the board to my computer as well as my phone. The picture aboce is how the data looks on my Samsung phone.
I developed an android app that is both compatible with AndroidOS and iOS capable devices. It was created in Unity3D and utilizes C# scripts. The objective of this game is to move the black player ball through the level utilizing 2 directions, right and left. Picking up the pink diamonds results in a boost to player score. The map the player goes through is randomly generated and is endless with asset recycling to minimize CPU usage. Scores are saved as well as the users highest score. This game was released to the google play store.
This program performs a Hough Transform on a user inputted image. The Hough Transform is a technique which is used to isolate features of a particular shape within an image by first performing edge detection and characterizing what is found. Eventually the algorithm will detect the most prominent line within the image and plot the data in Hough parameter space. This was coded in Matlab.
Using OpenGL and graphics shaders I created a bicycle that can be controlled with user input. When moving the bicycle wheels as well as pedals rotate according to the speed of the bike. The are the bicylce is in has randomly generated trees that when collided with the bike will result in a game over screen. This was programmed in C++ and various OpenGL API.
This program takes in a user provided image and will compute the histogram for that image. The histogram of an image is a graphical representation of the tonal distribution in a digital image. It plots the number of pixels for each tonal value and displays it in a graph format. This was programmed in Matlab.