• Model validation by working with AI models, evaluating their problem-solving processes and final answers to ensure accuracy and reliability.
• Failure testing by inducing logical errors in models to assess robustness and identify potential weaknesses.
• Evaluated models on various dimensions such as instruction following, verbosity, and correctness to enhance overall performance.

Clueless is a hackathon project designed to simplify outfit selection using AI. The app allows users to upload pictures of their wardrobe, categorize items with tags, and sync events from their calendar. Based on user preferences, weather conditions, and event types, Clueless intelligently suggests outfit combinations to streamline daily dressing decisions.

Built under a tight 2-day deadline, the prototype leverages Firebase for database management, React.js for the frontend, Flask (Python) for the backend, and GPT-2 as the AI model for outfit recommendations. While not fully functional yet, the project showcases the potential of AI-driven personal styling and wardrobe organization.

Jay is a data handling language designed to simplify and streamline data transformations. Key features include:

1. Smart Bundles
- Type-aware data grouping
- Automatic element-wise operations
- Flexible type matching and conversion

2. Intuitive Type Conversion
- Built-in conversion between common types
- Configurable conversion flags
- Safe handling of complex transformations

3. Procedure Pipelines
- Function composition using '>>' operator
- Clear data flow visualization
- Easy to chain and modify transformations

I plan to release a website for documentation with instructions on how to install.
I also plan to create libraries for popular open source languages and contribute to them.

Using: ANTLR, Kotlin, JUnit

My personal Portfolio website showcasing my skills and experience.

Used: React.js, github action pages for deployment

• Django-based multi-tiered system with a front end, Python script back end, and a PostgreSQL
relational database.
• Implemented systematic web scraping algorithms using Python, Requests, Regex, and BeautifulSoup to iteratively
extract UFC fighter and bout data from multiple webpage hierarchies.
• Designed and implemented ETL pipelines using Python and SQL, cleaning, and loading of data into the database,
ensuring accuracy and consistency through structured SQL database operations.

Future steps:
Incorporate data visualizations for each fighters data.
Deploy on cloud.
Predictive modelling

Tech stack:
Python (BeautifulSoup, Re, Requests)
Django
PostgreSQL

This was my final project in my scientific data analysis project, I conducted an in-depth analysis of crime statistics across Canada to uncover meaningful insights into societal safety and law enforcement dynamics. By examining various crime types, their frequencies, and geographic distributions, the study aims to reveal underlying patterns, trends, and regional discrepancies that may be influenced by socioeconomic and demographic factors.

At the heart of the project is the exploration of both reported and cleared crimes—the latter referring to cases that law enforcement has successfully resolved. This focus provides a dual perspective: while one aspect highlights the prevalence and distribution of criminal activity, the other offers an evaluative lens on the effectiveness of regional investigative efforts and the efficiency of justice systems across Canada.

The analysis is grounded in a meticulously curated dataset sourced from a reliable public data repository, encompassing multi-year records of crime statistics across various provinces and territories. Data cleaning techniques and statistical validation methods ensured the integrity and consistency of the dataset, setting a strong foundation for rigorous analysis.

Key components of the project include:

• Trend Identification: Unraveling temporal shifts in the rates of different crime types and observing the influence of seasonal or long-term socioeconomic changes on criminal behavior.

• Geospatial Analysis: Mapping crime occurrences to understand regional hotspots and variations, thereby providing insights into localized factors driving criminal trends.

• Cleared Crimes Evaluation: Assessing the proportion and distribution of cleared cases across different areas to evaluate the relative performance of law enforcement in resolving investigations.

• Data Visualization: Using Python’s data science tools—such as Pandas, NumPy, Matplotlib, and Seaborn—the project effectively communicates insights through clear, informative visualizations that facilitate policy discussion and public awareness.

By combining statistical analysis with insightful visualizations, this project not only underscores shifts in crime patterns but also contributes valuable information for policymakers, law enforcement agencies, and community stakeholders. Ultimately, the goal is to support informed decision-making aimed at boosting public safety and enhancing the efficiency of justice services

Tech stack:
Python Jupyter Notebook, Numpy, Pandas, Matplotlib