User's are able to input travel location, date, purpose, and length of stay through an initial input screen. If users are unsure about where they are travelling, they can also choose to chat with an AI agent to receive location recommendations.
The user will then move onto the chat page where they will receive an initial travel itinerary and have the option to chat with the agent to edit their itinerary.
When they are finished, they can view their itinerary on a final itinerary page. This page will also be stored in "itinerary history" on the app so user's can refer to old itineraries if they like.
The user will receive flight and hotel suggestions as well on the final itinerary page which is received from a Flask app that integrated a customized LangChain agent that I built. The app uses HTTP requests to get the necessary information.

Managing a group of 8 team members ranging from iOS, Android, backend, and marketing members. Organized biweekly sprints to develop new features for the app. In terms of backend development, integrated notifications using Firebase cloud messaging with the ExpressJS backend. Handleed the logic for notifications for departure as well as bus delays. Utilized Flask microservice to parse live bus data in order to notify for delays. Currently working on integrating AI to enhance route suggestions for users.

Upon downloading the app, users can enter their names and view a brief demo of how the app works.
After granting location permissions users will receive an outfit suggestion alongside a weather forecast.
Users can then select the wardrobe button to choose clothes from their closet to add to their current outfit on the main page. Users can also upload images of their clothes to their wardrobe.
Users will also have the option to pre-plan outfits for vacation with the vacation button at the bottom of the main page.

The goal setting app is powered by a React frontend and Flask backend. SQLAlchemy database is used to store user's goals. The backend uses sentiment analysis to get a numerical rating on the user's journal entries. Users may also evaluate their progress on their goals which is ran by OpenAI API supported agent.
As user's increase the progress of their goals, the plant that corresponds to their goals will grow accordingly.
User's can also chat with a virtual therapist bot which uses whisper and text-to-speech libraries in python.
A full demo and screenshots of the app can be seen on the frontend Github repo below.

The dataset used for this project are Amazon items.
Using a bag of words model, we stored each Amazon item as a vector of its description and features. We then stored all of our documents using an inverted index, the keys being the words and values being a list of all documents containing that word.
We used TF-IDF to correctly weighting the important terms. Common filler words like 'the' or 'a' are weighted less.
2 methods were used to return relevant documents to users. These would be cosine similarity and singular vector decomposition (SVD).
To further filter results, we implemented Rocchio's algorithm to provide more relevant results based on user's feedback.

In an attempt to mitigate the harmful effects reading triggering content may have on an individual's mental well being, Trigger Warning seeks to provide alerts when a website may contain triggering information. Website content is passed into OpenAI API and a response is returned which indicates if the website has the potential to be triggering.

To contribute to this open source project, I added several blocks of documentation providing examples of how to use the validation functions.

This website built in React displays an animated homepage, followed by a shopping page, cart page, checkout page, and return to shopping page.
Users are able to add items from the shopping page into their cart and in turn increase and decrease items in the cart as well.
The cart is also reflected in the checkout page and Email.js is used to send order requests.

Global Cornell Connection (GCC) is one of the leading business organizations within Cornell that develops competent leaders and business professionals.
Every recruitment cycle, GCC will receive roughly 100 applicants who submit applications through a Google form. To prepare for deliberations, the recruitment team will have to manually copy and paste all the applicant data into a PowerPoint for easy viewing.
This process proves to be tedious, thus, I have developed a Python script and Java code that can process applicant data and output a PowerPoint.

I created a Wordle plugin for the Jarvis open source project. Upon accessing the wordle command, users can play the word guessing game from their own computers' command line.

CUEat is a prototype of an app that is designed to allow Cornell students to post and share easy dorm recipes.
CUEat is prepopulated with several easy dorm recipes; however, students can also post their own easy recipes.
This app was developed with a team of 5 and I aided in the process of backend development. The backend code uses Python and SQLAlchemy. The backend code will accept new recipes and assign them to categories based on cuisine type, preparation time, and meal type.

The Path Planning project was completed with the project team CUAir at Cornell.
The purpose of the project was to help guide the aircraft to complete its 'airdrop' task. The plane must fly to 2 designated target locations to drop a water bottle and stay within competition boundaries in order to successfully complete this task.
For this project, I used Python to help develop an algorithm for how the plane would reach the target locations. This algorithm would choose waypoints that the plane would fly to that ensured the plane would stay within boundaries and reach the correct location.

Amyotrophic Lateral Sclerosis (ALS) disease is a neurodegenerative disease that affects all nerve cells in the body. Currently, there is no cure for ALS disease and the progression of ALS eventually leads to death.
This project utilizes past patient records and recognizes patterns in symptom onset in order to allow for early diagnosis of ALS disease. Early diagnosis allows for the possibility of prevention measures to be placed that can mitigate symptoms and even halt the development of ALS disease. Data mining algoirthm is used for pattern recognition in symptom onset. This project was completed with Princeton Pharametch.

From 2020 - 2021, I was an high school student researcher in Dr. Mark Albert's Biomedical Artificial Intelligence lab. Dr. Albert published a book titled Bridging Human Intelligence and Artificial Intelligence and I co-authored one of the chapters.
The chapter I wrote is titled "Early Visual Processing: A Computational Approach to Understanding Primary Visual Cortex". In this chapter, I highlight the similarity between early visual processing in the human brain in the primary visual cortex and Gabor wavelet codes. More about this chapter can be read in the hyperlink.

This Android app was also constructed under the guidance of Dr. Mark Albert's Biomedical Artificial Intelligence lab. The app utilizes simple gesture recognition to produce variance utterances. This is to aid the communication of those who have speech and motor disabilities like Cerebral Palsy.