January, 2022
Stock Market - Time Series analysis
Developed AR, MA, ARMA & ARIMA time series analysis models to analyze stock market performances of major stock market indexes like S&P 500, FTSE 100, & Nikkei 225.
Created comprehensive Jupyter notebook with stepwise analysis & explanation of all functions used.
Major learnings - Simplifying datasets, plotting data for analysis, splitting dataset into training & testing set, functional knowledge of time series analysis models like AR, MA, ARMA, ARIMA.
Libraries used - NumPy, Pandas, Seaborn, Matplotlib, Stats Model.
July, 2021
Company Database
Developed a corporate company database containing multiples tables comprising of data about employees, branches, clients, suppliers and sales.
Concepts used - Primary Key, Foreign Key, Unique Key, Views, Joins, Normalization, Transactions, subqueries, database design, concurreny & data models.
Tool used - MySQL Workbench
November, 2021
Visualising Citibike Trips
Analysed complex datasets from February 2018 to create data visualizations about average trip durations,
start times, stop times, most popular start & end stations in New York City.
Gained meaningful insights from data visualizations & published them into a dashboard.
Concepts used: data visualization, data analysis, importing data, data distribution, dashboard creation.
March, 2022
Executive Scorecard - Excel to Power BI
While working for Edgewell Personal Care (EPC) , I migrated the Executive Scorecard from Excel to Power BI by using DAX and Power Query
to provide more functionality, efficiency, and data visualization options to company stakeholders which helped them in making key decisions.
Major learnings - dashboards, reports, workbooks, datasets, vizualizations, and dataflows.
DAX functions used - FILTER, ALL, RELATED, TOTALYTD / TOTALQTD / TOTALMTD, CALCULATE, etc.
September, 2021
Nuclear Reactor Simulation
Created for CP213: Object Oriented Programming.
A simple nuclear reactor simulation. Given a starting temperature and control rods heights,
attempts to control the reactor over a period of time.
Major learnings - OOPS concepts (Data Abstraction, Encapsulation, Inheritance, Polymorphism),
Loops,
String handling,
Multithreading,
Exception handling,
Synchronisation,
Concurrent collection.
Framework used - Java Spring
August, 2021
BMI Calculator
Created for CP212: Windows Applications Programming
Developed a VBA application that helps users track their weight & food consumption.
Calculates BMI based on height & weight inputted & generates complete colorized dynamic BMI tables.
Added functionality to add user results to database using Microsoft Access.
Concepts used: VBA, modular programming, error handling, database management, macros, pivot tables.
April, 2023
Traffic Collisions in Toronto - Exploratory Analysis
The website is a Streamlit dashboard that provides insights on motor vehicle collisions in Toronto from 2006 to 2022. The dataset includes all traffic collisions where a person was either killed or seriously injured (KSI). The dashboard offers various interactive tools for data analysis, such as a map that shows the exact locations of KSI cases in Toronto categorized by the unique neighborhood ID setup by City of Toronto, and a slider to choose the year to analyze the trend of motor vehicle collisions. Additionally, it provides information on the number of KSI cases by hour of day and age group, helping users identify the peak hour and age group with the highest number of victims. The website is useful for people who are interested in analyzing the trends and patterns of motor vehicle collisions in Toronto.
Libraries used: Streamlit, Numpy, Pandas, Plotly, Plotly Express & Pydeck.
December, 2023
Optimizing Brain Tumor Detection with Convolutional Neural Networks (CNN)
The Business Analysis (BU425) group project addresses the inefficiencies in radiology, where 40% of a doctor's time is spent manually screening images. We developed a Convolutional Neural Network (CNN) model, leveraging transfer learning and training on 7,000 labeled MRI images, achieving 94% accuracy with the MobileNet model. This innovation automates brain tumor detection, reducing screening time by 20-30 minutes per patient. Estimated annual cost savings in Ontario exceed 46 million CAD. I played a pivotal role in researching & implementing different CNN libraries (ResNet, VGG19, Xception, MobileNet, DenseNet, Inception, VGG16), exploring market demand, and creating insightful visualizations using matplotlib and seaborn to compare accuracy, loss, and training time across six models. Our transformative solution aims for widespread adoption, emphasizing the need for trust in automated diagnostics. Received the award of "Best Project of BU425 for Fall 2023" by Professor Michael Pavlin.
Libraries used: TensorFlow (including Keras), Scikit-learn, Matplotlib, Seaborn, NumPy.
