The summer after my junior year of high school, I began working on a mobile application called Neighbor Ninja. The overarching purpose of it was to notify neighbors of suspicious activities in their communities. I used a crowdsourcing-based system and geolocation to achieve this. You can read more about it here. I spent June-August of that summer developing the application and was ready to launch before a friend notified me of a new app with a very similar purpose. They were backed by UC Berkeley and had a team of graduate students who ultimately created a tool Neighbor Ninja couldn’t match. This unfortunate event led me to abandon the project and it never reached the App Store.
Although the failure with Neighbor Ninja disappointed me and made all my hard work feel worthless, it helped me realize my true passion was for creating a product to better the populations’ daily lives in whatever way possible. Fast forward to my freshman year at the University of Michigan, I began brainstorming business ideas with my friend Alec. After a few days of freewriting and researching, we decided on developing a mobile application to simplify groups. We hated how difficult it was to truly connect with others in a group, like after joining a fraternity or club. You’d be added to a group chat with no way to actually document these connections without spending hours to find everyone on Facebook or Instagram. This idea served as the inspiration for Fuze, which offered the ability to create different social cards for different environments to easily share your networks with other people at the click of a button, as well as the ability to connect with everyone in group chats on different social networks with ease. After finding two more computer science students at U.C. Berkeley and U.T. Austin, we were able to build a minimum viable product within a month’s time. Unfortunately, our team would lose motivation after hitting bumps in the development cycle, and after a few weeks of work, I was left with the Xcode project and no partners to help me get the project to market. With little to no motivation left myself and finals looming around the corner, I lost focus on the project and also abandoned it.
Although I had a trail of abandoned projects behind me, I knew my passions were rooted in simplifying peoples’ lives and, as a result, I began working on another project with two of my friends, Junho Park and Sahil Patel, during the summer after my sophomore year. During the year, I had become a lot more interested in the stock market and started investing and following market trends. I knew applying my computer science knowledge to stock trading could simplify data analysis and make my time more efficient, which motivated the idea of analyzing sentiment data.
Three weeks into summer vacation, Junho, Sahil and I developed Sents and are excited to launch version 1. Sents is a tool designed to measure real-time general sentiments towards a specific stock. We periodically aggregate online messages about public companies and use a machine learning strategy to determine whether each individual message is bearish or bullish. We combine all these sentiments and run them through an algorithm that determines a final sentiment score which is displayed on the graph. We provide users the ability to compare this score against the actual changes in stock price with either side by side graphs or overlayed graphs. We currently support 200 tickers and take a sentiment score every 15 minutes.
Over the summer, we are all going to be interning at large companies but hope to continue adding new features to Sents. Looking forward, we hope to open up our data to the public by launching an open API, as well as setting up a user system so we can notify users on sentiment changes. We will also be diversifying our data set for calculating sentiment score, as well as retraining our model with new training data as it becomes available.
Looking back at projects I have worked on in the past, I believe this one actually came to fruition because of stronger team cohesion and more knowledge, which enabled me to tackle a harder project. I personally was in charge of the server and API, which was a fun and challenging job. I built the API using Flask and deployed it using Gunicorn and Nginx on AWS. Our front-end is running on Vue and Bulma, and we are using Tensorflow for the machine learning model.
I would appreciate if you could go fiddle around with Sents and leave UI or functionality feedback for us to implement in the coming months.