Rupal Gupta

With a mission to enable the growth of women in STEM fields and enhance their representation in technology, I built a Python-based web app for recommending mentors to women in tech. The web application allows users to register, log in, enter technical areas in which they require mentoring, search for mentors available on the platform, and connect with them via Twitter.

Technical stack: Vue.js, JavaScript, Bootstrap, HTML, CSS, Python, Flask, BeautifulSoup, NLTK, PostgreSQL, Git, Tweepy, Twitter API, and Microsoft Face API.

A deep learning project developed for the detection and localizing of house numbers in real images and videos. The project used convolutional neural networks along with a sliding window and non-maxima suppression techniques to build image pyramids. This performed multi-digit detection and evaluation of images at different scales and under different lighting conditions. The system was built using Keras and Python. The model produced an accuracy of 86.72% on the Google street view housing images dataset.

Developed a model that is capable of predicting unplanned hospital readmissions within 30-days of hospital discharge by using unstructured text data. Natural language processing was used to generate features to predict unplanned re-admissions from named entities in patient discharge notes. I used SparkNLP and PySpark to extract entities into an input matrix and build models using PyTorch and Scikit-learn.

Created interactive visualizations using D3 such as line charts, bar charts, heat maps, and a choropleth map for unemployment rates data. The development required data analysis and aggregation using JavaScript. The interactive functionalities built included creating tool-tips on hovering, hiding and displaying details in charts on mouse-out and mouse-over events, and displaying different information based on the drop-down selection.

Designed a machine learning-based trading agent by performing a time series analysis on stock data using Pandas, Python, and Scikit-learn. Devised technical indicators to evaluate the state of stock on each day, build a trading strategy learner using classification-based learner, and reinforcement learner- (Q-learning) based approaches and tested the learners on specific stock symbols and time periods.