Hi, I’m Vani! I’m currently pursuing a computer science degree, and I’m passionate about using CS to help improve people’s lives.
I’m interested in all sorts of tech — from 90 pound robots to detecting fake images via CNNs to making a portfolio (hi!!), I find CS exciting for all the possibilities and opportunities out there. Currently, I’m looking for an internship (Summer 2023) in either software engineering/development or data science so let me know if there are any interesting opportunities you think I’d enjoy by filling out the form at the end of the site!
Outside of tech, I enjoy writing articles for Font Femme, an all-girls STEM Magazine, reading & reviewing books , and drawing digitally .5 years +
Proficient with OpenCV, NumPy, Flask, Django, OpenAI, Pandas, TensorFlow, PyTorch
3 years +
Proficient with object-oriented programming (OOP) principles & Amateur with relational databases (mySQL, MariaDB).
~ 1 year
Amateur with querying & performing data analysis.
4 years +
Comfortable using git as a version control system on large-scale projects.
~ 1 year
Proficient using OOP principles to implement data structures.
~ 1 year
Proficient with TS + JS for full-stack development. Comfortable using React Native framework.
Project Euler is a website that contains a series of increasingly challenging mathematical/computer programming problems. Euler is designed to test a developer's understanding of the fundamental principles of a programming language. My goal is to solve one question (maybe more, maybe less) per day to learn new concepts, strengthen old ones, and eventually, to develop a more “optimized intuition."
In FRC, I worked with 120 lb robots and gained basic mechanical, electrical, and manufacturing skills. I also learned Java and basic motion control. During my time in robotics, I developed a vision system for automated robot sequencing via camera and got comfortable using git on a large project. Check out our 2019 and 2020 repo to see the cool stuff my robotics team did!
Round Up was a website I created with two friends for the MIT Hackathon. We created visualization tools to assist young students with mathematical concepts. All of the visualization tools were written in Javascript using the Processing library, and the website was written using HTML/CSS with certain elements from the Semantic UI framework.
This project trained custom TensorFlow networks for JeVois to detect corgis. The JeVois is a small smart machine vision quad-core camera. I started with a MobileNet network pre-trained on ImageNet and used transfer learning to fine-tune the network. Next I optimized for interference and converted it to TensorFlow Lite so it could run on the JeVois.
This version of Tic-Tac-Toe allows zero players, one player, or two players to play Tic Tac Toe. I wrote an "AI" player in Python that picked a move. I didn't want all games to end with the AI winning or a tie, so I worked on creating a "medium" intelligence. This project is written in Python with a focus on object-oriented design.
Palette locates the most dominant colors in an image using an automated version of the kmeans clustering alogirthm. I acomplished this using elbow plots and panda data frames. I estimated the number of clusters in a data set via the gap statistic. It is written in Python with a focus on object-oriented design.
Seam Carving Is a pair-programming project that I completed as the final project for CS60: Principles of Computer Science. It’s written in Java and demonstrates object-oriented and dynamic programming paradigms. The project was heavily rooted in computer vision and linear algebra, and its intent was to crop an image by removing the least “important” adjacent pixels in an image. This code is available on GitHub upon request. I recommend reading the wikipedia article on seam carving, it’s really interesting! Code is available upon request.
p-Deepfake is a project primarily written in Python that I worked on with 5 students from the 5C’s as a part of P-ai, a project-based organization exploring and democratizing the interdisciplinary applications of artificial intelligence.Deepfakes are forms of synethic media that can be used to spread fake news. They are extremely difficult to parse, and in this project, we analyzed feature data to determine the relationships between deepfake images and their real counterparts using machine learning algorithms SURF, SIFT, and ORB. This was more of an exploratory than analysis research project.
Moodboard is inspired by the heat-wave style maps Github uses to visualize a user's commits. It is a Flask application that uses the Spotify API and extracts metadata from a user's current music history to visualize their current mood. Specifically, I graphed the valence, which is audio feature data that measures how happy/sad a song is. (By design, this metric is quite subjective, but I thought it would be fun to visualize how a user's mood changes (or at least their song-choice) over time.)
Forward was a project I did with 4 other individuals for the Esri Intern Hackathon Project. We used GIS to design a web application that helped route individuals to the safest abortion clinic to them. Post Roe v Wade legislature was changing a lot and so we also built a live newsfeed (local to the USA) on abortion to ensure that users were making guaranteed safe decisions. We used ArcGIS Arcade, HTML, CSS, JS, Python, ArcGIS Pro, ArcGIS Online, and JavaScript APIs for ArcGIS Online to build this web application. We ended up being finalists and won Most Creative Project!
TODO is a VS code extension using java script, type script, svelte, and webviews that allows a user to maintain a list of tasks (TODOs) they wish to acomplish as a sidebar in VSCode.
This project used sentient analysis and predictive machine learning models to analyze the last 3 years of my Messenger data. Code is available upon request.