Projects

GitHub Repositories 3D Graph Visualization

Python, NumPy, Pandas, scikit-learn, NetworkX, OpenAI Embeddings, Tiktoken, Sigma.js

  • Visualizes GitHub repositories as an interactive 3D graph clustered by topic, labeled by language, and connected using cosine similarity.
  • Reveals relationships between projects through metadata extraction, vector embeddings, and graph-based visualization.

Generative AI for Improved Video Monitoring

Python, PyTorch, FFmpeg, OpenCV, YOLOv8, DeepSORT, 3D ResNet-18, Whisper (planned), Streamlit, Neo4j

  • AI-powered surveillance system for automating video monitoring, detecting anomalies in real time, and delivering actionable insights through an interactive dashboard.
  • Combines object tracking, spatiotemporal action recognition, and planned audio event detection to enable intelligent, real-time security analysis and reporting.

Transformer from Scratch with PyTorch

Python, PyTorch

  • Implemented the Transformer architecture entirely from scratch, covering self-attention, positional encoding, and encoder-decoder blocks without relying on pre-built libraries.
  • Provides a step-by-step, modular walkthrough of core components, masking strategies, and a basic training loop to demonstrate the internal workings of modern sequence-to-sequence models.

Face Recognition Attendance System

Python, TensorFlow, MySQL, OpenCV

  • Created algorithm for student identification and automated attendance tracking.
  • Utilized Python for image processing and data handling.

COVID-19 Detection using CNNs through Chest-Xray scans

Python, TensorFlow, Keras, Matplotlib

  • Developed a COVID-19 detection system from chest x-ray images using machine learning models like VGG16, achieving 60% accuracy.
  • Focused on optimizing model training and image preprocessing for better classification results.

Uber Ride Analysis

Python, NumPy, Pandas, Scikit-learn, Matplotlib

  • The project predicts Uber ride aspects using Logistic Regression and Random Forest models.
  • It compares model performance using test accuracy and cross-validation accuracy.

Adidas Sales Dashboard

Power BI

  • Designed a Power BI dashboard for Adidas sales, analyzing over 10,000 data points to uncover 15% revenue growth opportunities and improve regional sales insights.
  • Highlighted KPIs with a 35% average operating margin to enable data-driven decisions.