Roomy
Smart homes offer a set of basic device functionalities that users can use to build complex routines. But intelligent automatons require manual set up and normal interactions are handled with dashboards because voice assistants are flaky. Roomy is an AI powered smarter home proof of concept. We apply the tool forming and auto GPT concepts to build an intelligent assistant capable of handling complex, and opinionated interactions. System Design Roomy uses an orchestrator model to understand the users request and apply it to the most relevant command(s). If there are no commands that handle the user's request, the orchestrator calls the tool former. The tool former generates a new function building on the previous functions and it appends it to the set of functions for the orchestrator to call from. The tool former goes through a feedback cycle with a feedback agent until the tool former generates a function that compiles and meets the user's request (if possible). As roomy is used and it generates more functions it continuous to augment and improve it's abilities until it has a minimal amount of function creations. Power users can then write their own functions in python for the orchestrator to call, or they can edit python functions created by the tool former.
Magigram
A tool to help users create unique and meaningful pseudonyms. Inspired by Voldemort derived from the transformation of "Tom Marvolo Riddle" to "I am Lord Voldemort", I tried creating my own anagramatic pseudonym. It was a difficult and lengthy process. I created a tool to allow you to see all the valid word options, filter them, search through them, and select them. The product also includes a history section to allow users to explore multiple attempts simultaneously. Lastly, I created a transformer based inspired name generation algorithm to create a unique name constrained by the user's remaining letters. The model used is modifiable and users can weight different corpus to get a more Greek or Aztec inspired name for example.
Safe and Slug
UC Santa Cruz Hackathon 2024 submission. A progressive web app to facilitate safe events. We wanted to protect people at big events by empowering users to report health issues, violence, SA, or other issues discretely. The reports will be seen and addressed by event organizers who then have access to the reporter's phone number and disclosed medical information. With multiple user roles, we allow event organizers to view an event dashboard with all reports, all users, and users capable of helping in emergency situations. While the PWA is built in Next.js, tailwind, and daisy UI, and hosted on Vercel, we handle realtime data on a different server. Since each event manager will be looking at the dashboard, we need a better way of sending up to date traffic. We solve this by having a dedicated socket server to minimize traffic to the vercel server and to allow the server to update the dashboard in real time when necessary. The architecture is designed to be scalable and secure. Please view the dev post to see the design. To make the product easy to use, the website is responsive and downloadable as a PWA. We also made joining events seemless, by allowing users to scan a QR code with a link to the event that is generated on the event dashboard. There are a host of quality of life features as well, please reach out if you want to learn more.
TripBud
A Budget Trip Planner to help you plan trips within a budget and keep track of who paid for what!
Galacti-Pong
Circular pong inspired by: Paul Beltran
small side project to play around with web development and game design.
Scanslate
A travelers tool for language learning. It translates detected items to spanish and makes a flashcard with the picture for you.
Ambi~Air
An android app to help you save electricity, save the environment, and save your wallet! Ambi~Air is a project I built in highschool for my AP Computer Science A class final. This was my first project with Android, Firebase, Python or a RaspberryPi. It was built in 2 weeks and was a fun project. The RaspberryPi served as a weather station gathering weather data with an API and a temperature sensor. It sent the data to firebase using Python. The temperature however, was flawed. The RainbowHat's temperature sensor was directly above the CPU. I used a simple thermal transfer equation to estimate the added temperature, based on the temperature and distance of the CPU. On the other end, the Android phone running the app will see live updated temperatures obtained from Firebase. The app also does data manipulation to return a custom instruction alerting the user to open, or close windows when it would help the user get closer to their desired home temperature without using an AC or a heater.