top of page
ROSHNI'S PORTFOLIO
Admin Dashboard
Internal facing admin dashboard to track users
Smoking behavior prediction
Applies various machine learning algorithms (Decision Tree, KNN, Naive Bayes, SVM) to predict whether a person is a smoker based on health-related features. It covers data preprocessing, visualization, model comparison, and evaluation. Results show that a scaled SVM achieves the highest accuracy (~75.8%), though with higher computational cost.
AI Job Assistant
Conversational AI Job Assistant : A natural language interface to collect and understand job seekers' intent.
A simple command-line Python application that uses OpenAI’s GPT-4o model to extract job preferences from user input, fills in missing details via follow-up questions, and recommends matching jobs from a predefined job database.
Features:
Takes natural language input describing your ideal job.
Extracts key job attributes (role, location, salary, domain, company size, employment type) using OpenAI GPT-4o.
Saves extracted attributes in CSV format and loads them for validation.
Asks follow-up questions for any missing attributes.
Matches your preferences against a predefined job database.
Displays top 10 job matches with reasons for the match.
Cleans up temporary CSV files after execution.
Suppresses warnings for clean console output.
A simple command-line Python application that uses OpenAI’s GPT-4o model to extract job preferences from user input, fills in missing details via follow-up questions, and recommends matching jobs from a predefined job database.
Features:
Takes natural language input describing your ideal job.
Extracts key job attributes (role, location, salary, domain, company size, employment type) using OpenAI GPT-4o.
Saves extracted attributes in CSV format and loads them for validation.
Asks follow-up questions for any missing attributes.
Matches your preferences against a predefined job database.
Displays top 10 job matches with reasons for the match.
Cleans up temporary CSV files after execution.
Suppresses warnings for clean console output.
Stock 20-day SMA visualizer
#Fetching Financial data
Used yFinance to fetch financial data from Yahoo! Finance
#Matplotlib to plot the SMA 20 Data
20-day Simple Moving Average calculated by calculating the rolling mean of the closing stock prices over 20 days. Matplotlib plots the following data with AAPL in Blue in MSFT in Orange. The plot is equipped with a pointer to pop up whenever the graph line is clicked on
X axis spans the years. Y axis spans the SMA_20 prices.
#GUI to display financial data
GUI created using Tkinter is used to display historical stock data including Open prices, Close prices, High and Low prices, and Volume of stocks. This GUI features AAPL on the left side of the screen and MSFT on the right side, with both windows equipped with their own scrollbar.
#Running the program
Running the stock_prices.py file with simulatenously display the plotting interface along with the interface for the historical stock data. This is so because running this python file will run the data_GUI file along with it hence allowing for a more smooth user experience.
Used yFinance to fetch financial data from Yahoo! Finance
#Matplotlib to plot the SMA 20 Data
20-day Simple Moving Average calculated by calculating the rolling mean of the closing stock prices over 20 days. Matplotlib plots the following data with AAPL in Blue in MSFT in Orange. The plot is equipped with a pointer to pop up whenever the graph line is clicked on
X axis spans the years. Y axis spans the SMA_20 prices.
#GUI to display financial data
GUI created using Tkinter is used to display historical stock data including Open prices, Close prices, High and Low prices, and Volume of stocks. This GUI features AAPL on the left side of the screen and MSFT on the right side, with both windows equipped with their own scrollbar.
#Running the program
Running the stock_prices.py file with simulatenously display the plotting interface along with the interface for the historical stock data. This is so because running this python file will run the data_GUI file along with it hence allowing for a more smooth user experience.
Volatility Strategy on S&P 500 and Gold
Volatility strategy model utilizing historical data from the S&P 500 (SPY), characterized by high volatility, and Gold (GLD), exhibiting low volatility. The objective of this model was to explore the potential of diversifying a portfolio by incorporating these two asset classes.
Credit: The guidance and framework provided by CTG at the University of Wisconsin-Madison were instrumental in the development of this code.
Credit: The guidance and framework provided by CTG at the University of Wisconsin-Madison were instrumental in the development of this code.
S&P500 predictor
Utilized past decades data from S&P 500 stock prices to create the backend of an S&P 500 price predictor using a statistical linear ARIMA model. This model uses the past patterns in history to create a plot of what the forecast of future prices would look like.
Used: Python (Pandas, Matplotlib), CSV file
https://github.com/roshni-guha/CheeseHack-2024
Project made for Google CheeseHacks 2024.
Used: Python (Pandas, Matplotlib), CSV file
https://github.com/roshni-guha/CheeseHack-2024
Project made for Google CheeseHacks 2024.
Sustainability for Samsung
- As a team of five, we conceptualized and developed a sustainable cat food dispenser model utilizing repurposed Samsung TV boxes. Our objective was to promote sustainability and eco-friendliness by demonstrating the transformative potential of recycling in mitigating unnecessary cardboard wastage.
- Worked on the project for 3 months for almost 4 hours/week every week. I developed the detachable mechanism for the back and contributed to the development of the sliding mechanism of the dispenser.
- Developed a detailed instruction manual accessible through a QR code and filmed a pitch video explaining the project and objectives
Award: Most Innovative Award in Project Design Space by DIDI
- Worked on the project for 3 months for almost 4 hours/week every week. I developed the detachable mechanism for the back and contributed to the development of the sliding mechanism of the dispenser.
- Developed a detailed instruction manual accessible through a QR code and filmed a pitch video explaining the project and objectives
Award: Most Innovative Award in Project Design Space by DIDI
bottom of page