What if you could create data with an artificially intelligent neural network? I was astonished to find out that it was in fact possible for people to create data using neural nets. Data such as modified images where a horse is a zebra or an apple is an orange can be produced using algorithms that "think" intuitively. The power of GANs (Generative Adversarial Networks) allow users to create such data through an intuitive process where two neural nets are adversarially trying to reach a goal. I decided to test the potential of such models by starting a project where I would have to make a GAN that could turn an image of an apple into a realisitic image of an orange. This project allowed me to get started with learning about different ways neural nets could be used and it was cool to see something I made work as effectively as it did.
I began my career as a data scientist in 2018. Deep Learning specifically captured my interest because of how it performed accurately and how similar it was to the mapping of a human brain. In the last couple of months, I’ve created numerous image classification, object detection, generative adversarial network, and object detection neural nets. I wanted to see how neural nets would perform with text data. After doing some research and looking at how text is pre-processed and proccessed, I decided to create a simple network with keras/tensorflow 2.0 that could analyze the sentiment of a sentence the user provides. It was fun to learn about how text is transformed into numerical data and the specific architectures of the network which are optimal for such projects.
When my friend contacted me about doing an internship/software contracting job, I was extremely excited. I wanted to experience how it would be like to interact professionally with other people to get a serious task done by a deadline. IVAN (Intelligent Virtual Assisting Navigator) was an application I developed which could navigate through websites given a desired command from the user to pull useful information. Although it seems similar to web scraping, the system uses three Deep Learning models to interact with websites like a human. I earned $1.0k from this project but most importantly, I learned a lot and deep-dived into my journey as a data scientist.
My friends have been a huge influence on me in an entrepreneurial sense. I was inspired by them when they kept talking about creating their own companies and becoming entrepreneurs. I realized that security cameras in this era are either extremely expensive or not scalable. What if there was a product on the market that was cost-efficient, had AI incorporated in the system, and targeted public security? Thus, the idea of DeepNet began to grow. As of now, I am developing a prototype for the idea and I am also contacting investors as well so that I can make this soon to be company grow. By creating a product which most of the population can afford as well as targeting a whole new portion of the market, DeepNet has a lot of potential and is allowing me to gain a lot of technical/business experience as well.
This app was created when I realized that the process of signing up and creating clubs at Dublin High was extremely tedious and difficult for both the students and the staff. I started thinking of different ways I could help the school out and make clubs more accessible to students as well. That’s when I decided to create Club Central. After a couple days of brainstorming and polishing the idea, I decided to create an app using React Native and a firebase pipeline which could show all the clubs at Dublin High through the touch of your fingertips. This app has changed the entire club creation and sign-up process at Dublin High. With an expected user base of over 3000+ downloads, Club Central is one of my most popular ideas and revolutionized the way clubs work at my school.
Slider was the first mobile app I created which is currently available on the App and Play Store. When I began app development, I realized that the best way to learn was to start a project in which I could show other people who could give me feedback. I came up with a simple idea which I thought could be a fun first project for me to begin my app development career. A sliding puzzle game where users can select pictures from their gallery or take pictures using their camera allowed my users to enjoy a fun game whenever they wanted. With over 250+ downloads Slider was a fun and successful first project.
A lot of my friends were seniors before this idea came into my head. Seeing them struggle with paying for their undergrad and getting into the schools they wanted to showed me that the education system had to change. I got some inspiration from Uber and came up with an idea that I think will revolutionize the way education is received. Disciplina is a concept I came up with where users can find specialized “workshops” near them using an Uber-like google maps interface. The face-to-face interaction as well as the video/text interface feature makes this app much better than current competitors. Students can receive a cost-efficient personal education from qualified mentors looking to make quick side-hustle money.
I created this mobile application for 2018-2019 VRC Turning Point and 2019-2020 VRC Tower Takeover. This application contains multiple features that teams can use to scout different alliance partners. Some features this application has include a match predictor based on a FFNN trained from over 3000 past matches, a statistics displayer, a team information displayer, and a notes page. My team and I believed that by creating this application, we would be able to utilize its features to effectively scout alliance partners. This app is currently available on the App and Play store. The app currently has over 200+ downloads.
Programming is a tedious part of being on a VRC competition team because of how much you have to fine-tune the motor values. To help my team, I started to think of different ways we could quickly output autonomous code if we have a plan in our head. That’s when it hit me! What if you could output code by interacting with a program like a game? The Unity 5327 simulation which could effectively output code was born. The user starts the simulation and uses the arrow keys to move the robot around. After 15 seconds, the simulation outputs a c++ file which can be directly uploaded to a real robot. I also added an Artificially Intelligent A* search based agent that the user could play against. The judges at worlds said they had never seen anything like it before and it helped us win Create Award.