Participatory Budgeting Optimization – Cornell Tech
Worked with the NYC Civic Engagement Commission’s Chief Innovation Offer to boost civic engagement in participatory budgeting initiatives through novel data analysis and web-design techniques.
The project entailed analyzing budget proposal ideas, visualizing them on a map of NYC, and conducting sentiment analysis across boroughs of common community concerns.
Main methods employed include sentiment analysis, geographic mapping, & web-based visualization using NLTK, Folium, HTML, CSS, and JavaScript.
DataTool.tech
Your answer to visualizing your dataset and ensuring data quality.
DataTool.tech is an affordable and interactive AI-based online data quality platform which cleanses and visualizes uploaded datasets. Includes custom GPT with chat functionality.
Main methods employed include website design, prompt engineering, Python visualization techniques using ReactJS & Express.js.
Analyzing Gun Violence in Chicago
A 60-paged report exploring 60,000+ instances of gun-violence in Chicago through risk factors such as educational attainment, race, age-group, and geographical location.
Discovered upticks in gun violence during summer months and developed a regression model to predict one’s chances of experiencing gun-violence based on risk factors. Includes geographical distribution of gun crime, distributions of crime by demographic, and various binomial regression models to measure crime victim likelihood.
Main methods employed include machine learning regression models and evaluation techniques in hypothesis testing (T-Tests, F-Tests, KDEs) and model significance (MAE & RMSE), using Scikit-Learn, Statsmodels, Matplotlib, & Pandas.
Financial Data Science – Stock Market Trend Analysis
A data-driven approach to exploring the seasonality of the stock market.
The report contains technical indicator calculations, seasonality index charts, and vibrant heat maps showing price changes.
Main methods employed include stock and price visualizations using Seaborn.
Financial Data Science – Stock Market Recommender and Predictor
A homegrown model which visualizes tickers and corresponding technical indicators, and provides advice on stock trades based on trading strategy.
The model includes web scraped data, technical and fundamental indicators, price visualizations, homegrown trading strategies, and a backtesting algorithm to evaluate such strategies.
Main methods employed include technical indicator calculations, web scraping, price visualizations, and backtesting using Math, Pandas, yFinance, Matplotlib, & Backtester.
Data Science Consulting – PDF Parser
Worked with a college consultancy firm to automate data gathering from 9000+ student CommonApp application PDFs.
Main methods employed include tokenization, text parsing and extraction, and data frame design using Pandas, Regex, & PDF Query.
abhaygups 