blueZone
As Duke students, it’s extremely difficult to get candid opinions and thorough information on topics such as classes, majors, dorm culture, and dining options. Information crucial to the success of Duke students is often gate kept behind the walls of exclusive student groups. BlueZone aims to solve these problems; we plan to create a web app for students that leverages discussion boards and anonymous posting to facilitate honest conversations about all things Duke. The app will use a “groups and subgroups” structure to organize the discussion boards. Users can create posts specific to a group or subgroup as well as comment on and react to existing posts. There will be smart filtering on posts in the discussion groups so that users can sort by newest or most liked, for example.
We hope that BlueZone will foster greater conversations about topics Duke students are most curious about so that our peers can get clear information, especially in situations where they might be too afraid to ask. There is also an opportunity for this app to be extended to all universities, or even large companies where access to information and need for discussion is crucial. BlueZone will be open to anyone with a duke.edu email address.
Members. Joshua Boss, Zachary Lewiston, Grace Llewellyn, Arjun Rao, Alicia Steiman, Jodi Yeh
Building Sandbox World Playground for AI
Neural MMO is an open-source environment that simulates populations of agents in procedurally generated virtual worlds. It allows agents to forage for food, compete over resources, and struggle to survive. However, it lacks the incentivizing features that would encourage the creation of complex emergent social structures. We aim to provide these functionalities – specifically communication, productive tools, an exchange system for the agents for redistribution of resources, and many other means that encourages the societal formation in agents.
We hope that in the presence of a spectrum of new features, AI agents when learning to conquer the environment task, could form interesting societies as their strategies to optimize their survival in the environment. By providing agents with tools with progressing difficulties to craft and efficiency to produce, we hope to see agents learn how to organize social structures in an adaptive fashion to the changing means of production it has in order to maximize social welfare. However, we appreciate both the computational and algorithm challenges, but we do believe it is important to develop the groundwork in the first place in order for more future multi-agent learning research to come.
Members. Maverick Chung, Qiaoyi Fang, Lorne Zhang
ReCAPTURE
Currently, many websites use reCAPTCHA as an anti-fraud verification when users sign in or make a purchase. Users will have to complete tasks like labeling images of crosswalks or writing text from an image to prove that they are not a bot. Many times, these labeled data points are going to companies like Google who use the labeled data to train ML algorithms. However, websites that use reCAPTCHA capture no monetary value from this.
We propose Recapture, an alternative monetization mechanism for websites to capture value from reCAPTCHA. Instead of solely using reCAPTCHA for fraud prevention, websites will receive money from customers that require labeled data whenever their site visitors complete reCAPTCHA tasks. This paves the path towards a significantly more lucrative monetization mechanism than traditional ads, which have a CPM of $10 at best for YouTube/Google/Facebook ads. On the other hand, labeled data averages out to $25 per 1000 labeled images for the simplest classification tasks. This is a much better value prop to content sites that are at the mercy of Google ads for revenue.
Members. Che-wei Chang, Raymond Chen, Luke Qin, Jane Zhang
Conmigo App
- Conmigo, translates to with me in Spanish, con amigo = with a friend
- Hinting at Conmigo’s goal of blooming new friendships & connections across the Duke campus and different grade levels
- Goal: Cultivate new Duke student friendships over meals
- Students need a new way to socialize — especially during the post-pandemic stage
- Food is a popular commonality shared amongst college students
- We propose an iOS application to assist students in making new friends across campus, as we believe this will be an accessible and friendly way for shyer students to be introduced to new friends
- Our final deliverable will be a fully functional and user-friendly iOS application only for Duke students who seek to build relationships in an easy, low effort way
Members. Funda Eskalen, Ansh Nanda, Raffay Rana, Alexander Schiff, Zackary Schrage
Group Message Revamp
Our project, tentatively named Rally, intends to create a frictionless social event organizing experience for event-planners and participants. Our specific focus is on casual social events such as going out for drinks and dinner or bowling with a big group of friends. Too many plans fizzle out in the clutter of a group message, and it’s easy to forget the “with who”, “when”, and “where” regarding that concert you’re attending with friends from work. We hope to build an app that ensures more plans are seen through and fewer details are lost in conversation. We want Rally to make socialization easier, and we recognize that user feedback will be an absolutely essential part of our development process; we are excited to begin prototyping and putting our app to use along with our friends in order to inform further iteration.
Members. Matthew Finch, William Hayward, Joseph Lane, Sean Murphy
Patient Information App
An iOS application that allows healthcare providers (pediatricians) to have quick, easy access to basic health information about their patients. This information will include key patient demographics, medical history, lab results, and care plans. Health professionals will be able to view pertinent information quickly and hours of manual translation of information will be saved each day. At the end of the semester, we would like to have a mobile application that is in use at Burlington Pediatrics and has undergone several iterations based on user feedback.
Members. Justin Holmes, Ji Yun Hyo, Clara Lyra, Chris Yang
Smart Door on a Budget
Introducing Sentinel: Our goal is to allow students to control their door from anywhere, enabling them to unlock their door and grant access to whoever they choose with the use of a smart lock. Whether you forget your keys or a friend needs something from your room when you aren’t home, Sentinel will ensure your door can be opened whenever you need. Our hope is to create a functioning smart lock mechanism that can not only be opened with a tap of your phone using an NFC sticker, but can also connect to an iOS app which allows others to unlock your door during any given timeframe, and keep track of who has had access and unlocked your door.
Members. Isabel Garfinkel, Enoch Kuan, Raj Kundu, Sylvie Mason, Berkley Nahum, William Ramundo
STINGAR for Scalable and Secure Networking Architecture
STINGAR’s main goal is to provide security from network attacks by bad actors. One of the main ways that STINGAR accomplishes this goal is by using honeypots that act like real machines to catch network scan attacks that try to find vulnerabilities in the network. STINGAR then quickly blocks these IP addresses and shares them with other partners to help them stay secure before an attack even occurs.
The main problem that we are trying to solve is how we can modify existing honeypots such as Cowrie to make honeypots that are more tailored towards protecting higher education and industry networks and their respective group of partners. We hope that the final deliverables of this project are: (1) add onto and improve the existing STINGAR website/web app, (2) modify Cowrie to meet our specific needs and develop this honeypot, (3) if possible, develop a way to streamline a process of creating these new honeypots.
Members. Thomas Pratt, Jay Don Scott
