Find me here

Email Phone Github LinkedIn Letterboxd Minesweeper Scrabble
macdeath.xyz

Find me here

Email Phone Github LinkedIn Letterboxd Spotify Minesweeper Scrabble

All Projects

AI Threat Modeling Dashboard

Website
Python MongoDB FastAPI Next.js ReactJS

This is a full-stack AI threat modeling dashboard built to streamline security workflows for organizations.

Key Features

  • SSO Login: Admin authentication via secure Single Sign-On (SSO) with a dummy Admin Directory running locally.
  • Dashboard View: Interative view to monitor severities, tools, and assigned products and security champions.
  • Vulnerability Findings: Data integration with Defect Dojo to track findings and endpoints that can be filtered based on CWE, severity, scan tools, etc.
  • AI Threat Modeling: Upload Data Flow Diagrams (images) to get system analysis from locally running LLaVa model. Pass system analysis as context to locally running Mistral to get STRIDE-based threat modeling. Dynamic chat panel with context from chat history for follow up questions on uploaded DFDs. View previous threat-modelings chats in the history. Welcome modal with tutorial shown once per session per user.
  • AI chat assistant: Page-level AI assistant powered by ChromaDB for contextual help and guidance.
  • Product Assignment Workflow: Managers can assign projects directly to users under their hierarchy

Impact

Threat modeling analysis time was reduced from 60mins to 20mins for the security teams that tested it, and the tool is now being used by pilot team.

Asqi 1 Asqi 3 Asqi 2 Asqi 4

Snicket: A Book Exchange App

Github
ReactJS Firebase TypeScript

Social media inspired book marketplace app for connecting people through book exchanges and community-driven interactions.

Core Functionality

  • Marketplace Features: Users can scan ISBN barcodes or manually add books through our search feature to list their books for exchanges/lending.
  • Secure Direct DMs: In-app messaging for coordinating meetups and discussing books
  • Explore: Interact with other users through posts, comments and likes just like your regular social media
  • User Profile: Curate your reader profile by adding books to your profile that is visible to other users.

Impact

Connected over 200 people, with more than 30% of users moving from browsing to successfully exchanging books.

Snicket 1 Snicket 3 Snicket 2 Snicket 4 Snicket 5 Snicket 6 Snicket 7

Privacy Policy TLDR

Github
JavaScript CSS

This is a Chrome extension that summarizes company privacy policies into a short, readable format. It highlights the risk level, data collection, sharing, storage, and compliance info using Gemini's free API.

Key Features

  • Risk Level Classification: Categorizes data privacy risks as High, Moderate, Low.
  • Privacy Summary: Organised into data collection, sharing, storage and compliances.

Future Work

Currently exploring to extend the application Terms and Conditions summaries.

Privacy 1 Privacy 2

Symmetry Identification with Performance Constraints

Github
Python OCC

This project detects significant symmetry in 3D models (specifically STEP files) by analyzing the faces of the model. The tool works by identifying potential mirror planes that could reflect the shape symmetrically.

Research Context

This project was developed as part of my research in computational geometry and computer-aided design. The goal was to create an efficient algorithm for detecting symmetries in complex 3D models to optimize manufacturing processes and reduce computational overhead in CAD applications.

Algorithm Approach

  • Face Analysis: Extracts and analyzes geometric properties of all faces in the 3D model
  • Plane Generation: Generates candidate mirror planes using face centroids and normals
  • Symmetry Validation: Uses geometric hashing and spatial indexing for efficient symmetry verification
  • Performance Optimization: Implements early termination and hierarchical clustering to meet real-time constraints
Symmetry 1 Symmetry 2

Patient Health Journey Prediction

Github
Python

Developed a predictive model for analyzing patient health journeys and medication adherence patterns to help GoodRx improve patient outcomes. Using synthetic data of 4,363 hypertension patients, a Hidden Semi-Markov Model (HSMM) framework was adapted to classify and predict patient health journeys.The methodology combines k-means clustering, PCA, and HSMM to analyze unobservable health states through observed symptoms and medications

Faculty Research Trend Visualization

Github
Python PyTorch

This project presents a visualization framework for identifying research trends within university departments, focusing on Computer Science (CS) and Computational Science and Engineering (CSE) faculty at Georgia Tech. By using clustering algorithms and network graphs, the framework uncovers patterns, overlaps, and collaborations among faculty research areas.

Key Features

  • Research Clustering: Groups faculty based on research interests using clustering techniques.
  • Network Graphs: Visualizes relationships and interdisciplinary connections.
  • Pattern Analysis: Identifies key thematic areas and emerging trends.
  • Reusable Framework: Can be adapted for other academic departments.